1_data_preparation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1_data_preparation.ipynb\n",
    "\n",
    "Converts 'open-high-low-close' (OHLC) price data into a format suitable for machine learning.  Target labels are generated according to (potential) profitable trading conditions being met.  Extra data features are created, price and volume data are centralised around their moving averages, whilst time and day labels are separated out and one-hot encoded.  \n",
    "\n",
    "The working dataframe is then transformed into a tensor of dimensions [batch_size, steps, channels] by using a rolling window method.  Appropriate category labels are selected to match the last step in each window, and transformed into a tensors of dimensions [batch_size, categories]."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from imports.noise_generator import NoiseGenerator\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import time\n",
    "import pickle\n",
    "\n",
    "pd.set_option('display.max_columns', 50)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Exploration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAj4AAAG1CAYAAAD5rf4qAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8g+/7EAAAACXBIWXMAAA9hAAAPYQGoP6dpAADpwUlEQVR4nOydd3QbVdrGf3JNc5zeK+kJaUCABEgChBI62QWWGlg+YJeyLGUpC7uh97KU0GuoSyhZSoCEkrgXufcud7lbli1bdb4/BgsrtiWNRrIkZ55zdGzrvs9bxs9cX1/duVclCIKAAgUKFChQoEDBYYAQfyegQIECBQoUKFAwVFAGPgoUKFCgQIGCwwbKwEeBAgUKFChQcNhAGfgoUKBAgQIFCg4bKAMfBQoUKFCgQMFhA2Xgo0CBAgUKFCg4bKAMfBQoUKBAgQIFhw2UgY8CBQoUKFCg4LBBmL8TCCTYbDbq6uqIiopCpVL5Ox0FChQoUKBAgRsQBAG9Xs+MGTMICXE+p6MMfPqgrq6O2bNn+zsNBQoUKFCgQIEHqK6uZtasWU5tlIFPH0RFRQHihRs7dmy/9uLiYhYvXjwo31m72Wxm3759nH766YSHh3snYT/B1XUIprhyfXrCl8Jx11bRpghFm/L4ijZ9B0Wb8viuOB0dHcyePdv+d9wZlIFPH/R+vDV27NgBBz7R0dEDvu9Ou9lsZtSoUYwdOzbob2BX1yGY4sr16QlfCsddW0WbIhRtyuMr2vQdFG3K47vLcWeZirK4WQJGjx4tq324wF91+iKuXJ+e8KVw3LVVtClC0aY8vqJN30HRpjy+N+tQBj4SUF9fL6t9uMBfdfoirlyfnvClcNy1VbQpQtGmPL6iTd9B0aY8vjfrUD7qkoAlS5bIah8u8Fedvogr16cnfCkcd20VbYpQtCmP701t2mw2TCYT8+bNo6enZ1A7Z+1ms5mwsDB6enqwWq1u5xaIcHUdgimuXJ+e8OfPn4/NZnP5xJY7UAY+EpCens7GjRs9bh8u8Fedvogr16cnfCkcd20VbYpQtCmP7y1tmkwmKioqsNls9PT0MGLEiEH9OGsXBIFp06ZRXV0d9FuMuLoOwRRXrk9P+D09PTQ1NTF//nwiIiI8jg2gEgRBkOVhGKGjo4Po6Gh0Op3XF4OZzWb27t3LWWedFfSL9BQMLyjaVOBNCIJAVVUVZrPZrT1VnMFms9HZ2cmYMWO88p++guBF7z574eHhzJkzp99AWNLfb0GBHTqdTgAEnU43YPvBgwed8p21m0wmYc+ePYLJZJKVYyDA1XUIprhyfXrCl8Jx11bRpghFm/L43tCmyWQS8vPzhfb2dkEQBKGjo8OpH2ftVqtVaGtrE6xWq9t5BSpcXYdgiivXpyf8jo4Oob29XcjPzx+wr3L197svlBmfPnA1YjSbzU7/I3bWPpz+q3Z1HYIprlyfnvClcNy1VbQpQtGmPL43tNnT00NFRQXz5s1j5MiRLtdlOGu32Wx0dHQwduzYoJ/x8db6lECIK9enJ3ybzYbRaESj0TB//vx+H5VJmfEJbiUNMVJTU2W1Dxf4q05fxJXr0xO+FI67too2RSjalMf3pjZ7P4owGAxO7Vy1Dxf4q05fxJXr0xO+wWDw2jovZeAjAcqTMyKUJ2fk8ZWnunwHRZvy+L7QpqtFrP5Y8OsP+KtOX8SV69MTvjfrUAY+EqDsRyFC2Y9CHl/ZK8V3ULQpj+8LbZrNZlntwwX+qtMXceX69ITvzTqUgY8EKDuQilB2IJXHV3bH9R0Ubcrj+0KbrtZyBMranc2bN/P3v//dZ/79Vae7cTUaDSqViszMTK/59Cbfm9cvMBTnZ+zcuZPly5ezbt06AFpbW4mPj8disRATEwNATEwMRqOR9PR0amtrKS0tpaCggKamJpKTkzEajeTk5Nhtu7u7SUlJoaGhgaKiIkpLSzEYDKSnp9PV1eXg12w2k5CQQHNzM3l5eZSXl1NdXU1mZiY6nc7B1mazERsbS3t7O1lZWVRVVaHRaMjNzaWlpWXAvDs7O0lLS3PIu7GxkeTkZHp6ehxse3p67HkXFhZSUlJCXV0daWlpdHZ2EhMTQ2hoqEPeLS0t5ObmUlFRQVVVFVlZWfa8BUFw+KrT6ex5V1RU2PNOSEjAbDYPmHddXR0lJSXU1NTQ0NBASkrKgHknJyfT2NhIQUEBpaWl1NbWOuTda2uxWIiPj6elpYXKyko0Go097/b2dmJjY7HZbA4cnU5HZmYm1dXVlJeXk5eXR3NzMwUFBf3y7urqQq1WU19fT3FxMUVFRfa8u7u7HXRiNBpJTk6mqanJIe/09HT0er2DbW/era2t5OTkoNFoqKysJDs7G4PBQFxcXL+8Ozo6yMjIoKWlhbKyMnveiYmJmEwmu61Wq8VgMKBWq9Fqtfa8tVotqampGAyGfvdCUlISTU1N5OfnU1ZWRk1NDRkZGej1egdbq9VKXFwcbW1tZGdnU1lZac+7ra2NuLg4rFarA0ev15ORkUFNTQ1lZWXk5+fT1NREUlISRqPRwdZgMJCamopWq6W2tpbi4mK0Wi1qtbpf3iaTicTERPu91jfvjo6OfvdaXFwc7e3t9rw1Gg05OTn9+ojee0Kv1zvtI/r6P7SPKC4upr6+HrVaTVdXl8Pv3pM+oru7W3If0dra6nYfodPpBuwjetf+dHV12ffxsVgsdHd3YzQaMRqNGAwGLBYLer0eAL1ejyAI6PV6LBYLBoMBo9GIyWTCYrFgtVrp7OzEZrM5cKxWK11dXZhMJnp6euju7sZsNttj97XtfTTebDbT3d1NT08PJpOJrq4urFarfYPE3lw6Ozud5t2bby+nN2+TyYTRaLTn0pt3d3f3oHn39PR4nPeh17A3795r2Pv9YHlbrVYMBgNTp06loqKCBQsWOOQ90PW2Wq398j40l8HyNhgM/fwO9LsfSCddXV2AuL7s0D6isLAQtyH5mbJhDFePw2VmZjrlO2sfTo8Mu7oOwRRXrk9P+FI47toq2hShaFMe3xva7O7uFvLz84Xu7m5BEAShq6vLqR9n7UP5OPumTZuEW2+9VRLHaDS6bevqOvgKvogr16cn/K6urn7a6gspj7MrMz4SMG/ePFntwwX+qtMXceX69IQvheOuraJNEYo25fF9oc3w8HAMJsugLwshTtu7TVan7c5egsTdWiwWCzfffDPR0dFMmjSJf/3rXw4+5s2bx8MPP8xVV13F2LFjuf766wH44osvWLFiBZGRkcybN49nn33Wznn55Zc58sgj7bsN79mzB5VKxWuvvWa32bJlC/fffz8ADzzwAGvWrOGDDz5g3rx5REdH86c//ck+6zEQ3nvvPcaNG8ePP/7IsmXLGDNmDGeeeSb19fX2uDabjYceeohZs2YRGRnJmjVr+OGHH+w+Dv2oq62tjcsvv5zJkyczcuRIFi1axLvvvgtAREQE1dXVXHzxxYwbN44JEyZw/vnno9Fo3LrOnuy8LHe35r5QjqyQgKysLKfbubtqHy7wV52+iCvXpyd8KRx3bRVtilC0KY/vC222dnRx3NPxkvLwFvIfOoNREe7/mXv//fe59tprSUlJQa1Wc/311zNnzhyuu+46u80zzzzDv//9b3bs2AFAWloaF198MQ888ACXXHIJCQkJ3HjjjUycOJGrr76aTZs28be//Y3q6mrmz5/PwYMHmTRpEgcOHOAvf/kLZrOZxMRE7rnnHnuMsrIy9uzZw7fffktbWxsXX3wxTzzxBI8++uiguRsMBp555hk++OADQkJCuOKKK7jzzjt57bXXiIqK4oUXXuDZZ5/l9ddfZ+3atbzzzjucd9555OXlsWjRon7+/vWvf5Gfn8/333/PpEmTKC0ttX9c19HRwRlnnMH69euJjY0lLCyMRx55hDPPPJPs7GyXg5Tu7m6ioqLc/r30cry1V5aygWEfuNoASRAEp/sIOGsfTpvEuboOwRRXrk9P+FI47toq2hShaFMe3xva7N3AsHeTuS6jmRU79knKw1uQMvDZvHkzjY2N5OXl2eu65557+Prrr8nPzwfEGZ+1a9fy1Vdf2XmXX345TU1N7Nv3e4133XUX3333HXl5eQiCwOTJk3n11Ve56KKLWLt2LZdccgkvvPAC9fX1xMfHc/LJJ9Pe3s6oUaN44IEHePrpp9FqtfbBwV133UVMTAxJSUkD5v7ee+9xzTXXUFpayoIFCwB45ZVXeOihh6ivr0elUjFz5kxuuukm/vnPf9p5xx57LOvWrWPnzp32jQEzMjJYs2YN5513HpMmTeKdd97pF++DDz7g0UcfpaCgwH6tTCYT48aNY8+ePZx++ulOr7Wn2jQajQ7a6gspGxgqMz4SEBsb6/Q/HFftwwX+qtMXceX69IQvheOuraJNEYo25fF9oU2rsZv8h84YtF2v7yQqasyAbTabDX2HnqixUR491TMyPFSS/fHHH+/wB3n9+vU8++yzWK1WQkNFX8ccc4wDp6CggPPPP9/hvRNOOIH//Oc/dt7GjRvZv38/p512Gvn5+dx444089dRTFBYWcvDgQdatW8eoUaPs/Hnz5jnMiEyfPp3GxkanuY8aNco+6OnL6ezsRBAE6urqOOGEE/rlmZWVNaC/v/71r/zhD38gPT2d008/nQsuuIANGzYAoFarKS0t7Tdr09PTQ1lZmdM8ATo7OyXP+HR2dnrtHzNl4CMBJ510kqz24QJ/1emLuHJ9esKXwnHXVtGmCEWb8vi+0GZUVJTT/+5HTogetN1ms2GJCGVURFjAPPbuyTYBmzdv5o033iA2Npa1a9cyduxYNm7cyIEDBzh48CCbNm1ysD/0D7xKpcJmszmNMRBHEATGjBnjdH3QYNi6dSuVlZXs3buX/fv3c+qpp3LTTTfxzDPPYDQaOfroo/noo4/68SZPnuzS95gxAw90XXGMRqNk3kAIDCUFCWJjY2W1Dxf4q05fxJXr0xO+FI67too2RSjalMf3hTY7OztltQ8lkpOTHX5OSkpi0aJF9tmegbBs2TLi4x3XMMXHx7N48WI7b9OmTeTn57N79242b94MiIOhn376ifj4ePt7vkBnZydjx45lxowZA+a5fPnyQbmTJ09m+/btfPjhh/znP//hjTfeAGDFihWUlJQwZcoUFi5c6PCKjo52KydP6vAWlIGPBKxevVpW+3CBv+r0RVy5Pj3hS+G4a6toU4SiTXl8X2hz5MiRstqHElVVVdx+++0UFRXxySef8NJLL3Hrrbc65dxxxx38/PPPPPzwwxQXF/P+++/z8ssvc+edd9ptVq1axfjx4/n4448dBj579uzBaDT2+wjKm+i9vv/4xz948skn+e9//0tRURH33HMPmZmZg9b373//m//973+UlpaSl5fHt99+y7JlywDYvn07kyZN4vzzzyc2NpaKigoOHDjA3/72N2pqatzOyZM6vAFl4CMBrh7Vc/dRvmCHv+r0RVy5Pj3hS+G4a6toU4SiTXl8X2jTZDLJah9KXHXVVXR3d3Psscdy0003ceutt9ofWR8MRx11FJ999hmffvopRx55JP/+97956KGHuPrqq+02KpWKDRs2oFKpOPHEEwFxMDR27FiOOeYYn+443nt9//a3v3H77bdzxx13sHLlSn744Qe+/vrrAZ/oAvHx8XvvvZdVq1axceNGQkND+fTTTwEICwsjJiaGOXPmsG3bNpYtW8a1115LT0+Py4XFfXPypA5vQHmqqw/sq8L3P83Y0f1Hl21tbYwfP35QvrN2q9VKfJWZ9Rf9LeifnKmqqmLOnDnDIq5cn57wpXDctXVl56x9OD3VpWhTHt8b2jz0qS6j0UhkZOSgfpy122w2Ojo6GDt2bMCs8fEUrq5DMMWV69MTvtFoRBAE5akun+HnhyCy/2K7wYc8rttDgRNU4djazoUpi+Vk53f0bu0+HOLK9ekJXwrHXVtXdv76nQ01FG3K4/tCmwoUBBqUgc9AWH4BjOq/AVN7ezvjxo0blOasXajPIrS5GNW+++CK3d7J00/oPS9lOMSV69MTvhSOu7au7Pz1OxtqKNqUx/eFNl09jeSqfbjAX3X6Iq5cn57wbTab1/bKUgY+A+H8l2GAqTJrSwtMnDgozVm7pT6f0DdOJKR0HxR9D0u2ei3docb06dOHTVy5Pj3hS+G4a+vKzl+/s6GGok15fF9o09XHp8H+8aq78Fedvogr16cn/PDwcCwWi6y4vQjuD02HGEVFRZ63T1pE6ZTfBjvf3wXmbi9mNrRwdR2CKa5cn57wpXDctZWlzWEERZvy+L7QZk9Pj6z24QJ/1emLuHJ9esL3Zh3K4uY+cLU4ymw2Ox2pOms3m838+M1XnK15AFVHLWy6G07+54C2gQ5X1yGY4sr16QlfCsddW7naHC6LmxVtyuN7Q5uHLm622WxOFyY7ax9Oi5tdXYdgiivXpyd8m82GyWTyyuLm4FaSl7Bz506WL1/OunXrAGhtbSU+Ph6LxUJMTAwAMTExHDhwgPT0dGprayktLaWgoICmpiaSk5MxGo188skndtvu7m5SUlJoaGigqKiI0tJS9EYrJQv+DIAt9jloKSMmJgaz2UxCQgLNzc3k5eVRXl5OdXU1mZmZ6HQ6hxxsNhuxsbG0t7eTlZVFVVUVGo2G3NxcWlpaBsy7s7OTtLQ0h7wbGxtJTk6mp6fHwbanp8eed2FhISUlJdTV1ZGWlkZnZycxMTEkJiY65N3S0kJubi4VFRVUVVWRlZVlz1sQBIevOp3OnndFRYU974SEBMxm84B519XVUVJSwtdff01DQwMpKSkD5p2cnExjYyMFBQWUlpZSW1vrkHevrcViIT4+npaWFvbs2YNGo7Hn3d7eTmxsLDabzYGj0+nIzMykurqa8vJy8vLyaG5u5r///W+/vLu6ulCr1dTX11NcXExRUZE97+7ubgedGI1GkpOTaWpqcsg7PT0dvV7vYNubd2trKzk5OWg0GiorK8nOzuaXX34hLi6uX94dHR1kZGTwww8/UFZWZs87MTERk8lkt9VqtRgMBtRqNVqt1p63VqslNTUVg8Hg4NdoNJKUlERTUxP5+fmUlZVRU1NDRkYGer3ewdZqtRIXF0dbWxvZ2dlUVlba825rayMuLg6r1erA0ev1ZGRkUFNTQ1lZGfn5+TQ1NZGUlITRaHSwNRgMpKamotVq+eabbyguLkar1aJWq/vlbTKZSExMtN9rffPu6Ojod6/FxcXR3t5uz1uj0ZCTk9Ovj+i9J/R6vdM+oq//Q/uI4uJi6uvrUavVdHV1OfzuPekjfv31V8l9xI8//uh2H7F///4B+4jU1FRAXANks9loa2vDYrHQ3d2N0WjEaDRiMBiwWCzo9Xq6urrQ6/UIgoBer8disWAwGDAajZhMJiwWC1arlc7OTvEIi992Idbr9VitVrq6ujCZTPT09NDd3Y3ZbLbH7mtrs9no7OzEbDbT3d1NT08PJpOJrq4urFarg60gCHR2djrNuzffXk5v3iaTCaPRaM+lN+/W1tZB8+7p6fF63r3XsKOjw2neVqvVnnffa+jsend0dPTL+9BcBsvbYDDQ2dnZL+9Df/cD6aR3XVlqamq/PqKwsBB3ocz49IGUEaNU2P+r3rqV8P9eAmW/wMLT4PLd4IdDFRUo6MVwmvFR4H8cOuMjB8NpxkeBfDjTljLj4yP0/tfjaTsgDnK2Pg0h4VC6H4r2eim7oYNbdQZJXLk+PeFL4bhr6xVtDgMo2pTH94U2XZ0T5ck5UsEIf9Xpi7hyfXrC92YdysBHAo466ihZ7XZMWggbbhG///4eMBlkZja0cLvOIIgr16cnfCkcd229ps0gh6JNeXxfaLPvqeOetA8X+KtOX8SV69MTvjfrUAY+EuDVJ2c23gljZ4GuCuKel5nZ0EJ5ckYeX3mqy3dQtCmPrzzVNTA0Gg0qlYrMzEyPfQTCU10PPPAAa9as8apPOfyrr76aCy64YEhi9oUy8JEAr+6VEjEaznxc/D7+P9BS5nliQwxlrxR5fGUfH99B0aY8vrKPj+8QCPv43Hnnnfz8889e9TkY6uvrueyyy1i8eDEhISH8/e9/l8T3JKa7UAY+EuD13XGXnQsLTgWrCb6/G4JknbmyO648vrJzs++gaFMeX9m52Xfw587NgiBgsVgYM2YME51swivFpysYjUYmT57M/fffz+rVqyXzPYnpLpSBjwR4/TwklQrOehpCI8SFzoXfychu6KCchySPr5zV5Tso2pTH94U2BZsNTF2DvgRjp9N2zAbn7c5eEv+ZtNlsPPXUUyxcuJDIyEjmzJnDo48+OqDtwYMHOfbYY4mMjGT69Oncc889DjsLf/7556xcuZKRI0cyceJEzjrrLIfB4ltvvcWyZcsYMWIES5cu5ZVXXnErx96P3T799FM2bNjAiBEjOPLIIzl48KDd5sCBA6hUKr7//nvWr19PZGQkcXFxA37U9c4777BixQp7HTfffLO9rb29nf/7v/9j8uTJjB07llNOOYWsrCzceRh83rx5vPDCC1x11VVER0c7tB3Kf+aZZ5g+fToTJ07kpptuwmw29/PnzQfQlSMrJMDZyezutA+IiQtgw98g9hn44R5YcApEBPZiP4/qDNC4cn16wpfCcdfWJ9oMQijalMf3hTbDBBM8NnfQ9pFOuCHAOLczGgD/rBOXFbiJe++9lzfffJPnn3+eE088kfr6+gH3h6mtreWss87i6quvZteuXRQWFnLdddcxYsQIHnjgAerr67n00kt56qmnuPDCC9Hr9Rw4cMD+x/ujjz7i3//+Ny+//DJr164lIyOD6667jtGjR7N9+3a3cv3HP/7Bf/7zH5YvX85zzz3HueeeS0VFhcOMzj333MMTTzzB4sWLGT9+PAcOHHDw8eqrr3L77bfzxBNPsHXrVnQ6HfHx8fb2iy66iJEjR/L9998THR3N66+/zqmnnkpeXh4jRzr7zTlHWNjvQ49ff/2V6dOn8+uvv1JaWsoll1zCmjVruO666/pxvPUPhjLjIwGVlZWy2gfFSXdA9GzQVUPss575GEJ4XGcAxpXr0xO+FI67tj7TZpBB0aY8vi+0aTL1/+89EKHX63nhhRd46qmn2L59OwsWLODEE0/k//7v//rZvvLKK8yePZuXX36ZpUuXcsEFF/Dggw/y7LPPYrPZqK+vx2KxsG3bNubNm8fKlSv585//zJgxYwDYsWMHzz77LNu2bWP+/Pls27aN2267jddff93tfG+++Wb+8Ic/sGzZMl599VWio6N5++23HWweeughNm7cyIIFC5gwYUI/H4888gh33HEHt956K4sXL2bdunX2tThxcXGkpKSwe/dujjnmGBYtWsQzzzzDuHHj+OyzzyRc2f4wmUz278ePH2+/jueccw5nn332gOuQ+nLkQpnxkYBVq1bJah8UEaPgzCfgv5dDwouw+lLxkfcAhcd1BmBcuT494UvhuGvrM20GGRRtyuP7Qpsjx04QZ14GgcsjK/R6xkZFebaBYbj7s+cFBQUYjUZOPfVUt2zXr1/vcFr4CSecQGdnJzU1NaxevZpTTz2VlStXcsYZZ3D66aezbds2Ro8eTVdXF2VlZVx77bUOsxoWi6XfR0LOsH79evv3YWFhHHPMMRQUFDjYHHPMMYPOzDQ2NlJXVzdovVlZWXR2dvZbE9Td3U11dbXbeQ6EvjmtWLGC0NBQ+8/Tp08nJydnQI63Bj/KjI8EpKSkyGp3iqVnw8Itvy10viugFzrLqjPA4sr16QlfCsddW59qM4igaFMe3xfa7DIYxI+bBnl1mXHaTvgo5+3OXhJ2xZfz0c2hCA0NZf/+/Xz//fcsX76cl156iaVLl1JRUUFnZycAb775JpmZmfZXbm4uSUlJXssBsA+0BoKrejs7O5k+fbpDjpmZmRQVFXHjjTfKyqtvToc+raVSqQZcyOzNBwiCZuAzb948VCpVv9dNN90EiGcNXXnllUybNo3Ro0dz1FFH8cUXX3g1h40bN8pqdwqVCrY+JS50LvsZCr/13JePIavOAIsr16cnfCkcd219qs0ggqJNeXxfaDMqKkpW+1Bh0aJFjBw50q3HvZctW0ZiYqLDgtv4+HiioqKYNWsWIP4BP+GEE3jwwQfJyMggMjKSr776iqlTpzJjxgzKy8tZuHChw2v+/Plu59t3kGSxWEhLS2PZsmX97Aa7vlFRUcybN2/Qeo866ii0Wi1hYWH98pw3b57beQ4Weyg4gyFoBj6pqanU19fbX/v37wfExVcAV111FUVFRXz99dfk5OSwbds2Lr74YjIyMryWg8+PBZi4AE64Vfz+h3vFpxICEMqxAPL4ypEVvoOiTXn8w/nIihEjRnD33Xdz1113sWvXLsrKykhKSuq3bgbgxhtvpLq6mltuuYXCwkL+97//sWPHDm6//XZCQkJITk7mscceQ61WU1VVxZdffklTU5N9YPLggw/y+OOP8+KLL1JcXExOTg7vvvsuzz33nNv57ty5k6+++orCwkJuuukm2tra+POf/9zPztn1feCBB3j22Wd58cUXKSkpIT09nZdeegmALVu2sH79ei644AL27duHRqMhISGB++67z+EJMmfonSXq7OykqamJzMxM8vPz/X5kBUKQ4tZbbxUWLFgg2Gw2QRAEYfTo0cKuXbscbCZMmCC8+eabbvvU6XQCIOh0ugHbu7u7nfKdtZtMJmHPnj2CyWRynoSxSxCeO1IQdowVhJ8edJmzP+DqOgRTXLk+PeFL4bhrOyTaDAIo2pTH94Y2u7u7hfz8fHu71Wp16sdZu9VqFdra2lz68BasVqvwyCOPCHPnzhXCw8OFOXPmCI899phQUVEhAEJGRobd9sCBA8K6deuEiIgIYdq0acLdd98tmM1mQRAEIT8/XzjjjDOEyZMnC5GRkcLixYuFF1980SHWRx99JKxZs0aIiIgQxo8fL2zcuFH48ssvXebYm8vHH38sHHvssUJERISwfPly4ZdffrHb/PrrrwLQ79rt2LFDWL16tYO/1157TViyZIkQHh4uTJ8+XbjlllvsbR0dHcItt9wizJgxQwgPDxdmz54tXH755YJGo3HregL9XnPnzrXntH37duH888934Nx6663Cpk2b+vmyWq39tNUXrv5+90VQns5uMpmYMWMGt99+O//85z8BOP3004mIiGDXrl32VefXXnstWVlZLFw48EJho9GI0Wi0/9zR0cHs2bNpbm4e8HRXtVrNMcccM2heztrNZjP79+/ntNNOc7kDpapoL2GfX4UQEo7l+liYGFgLnV1dh2CKK9enJ3wpHHdth0qbgQ5Fm/L43tBmT08P1dXVzJs3jxEjRmAwGJyes+SsXRAE9Ho9UVFRDguJgxGuroO70Gg0LFiwgLS0NLeOn/BWXG/69IRvMBgICQlBo9Ewe/bsAU9nnzRpklunswflwOezzz7jsssuo6qqihkzZgDiRkuXXHIJ+/btIywsjFGjRrF7925OP/30Qf088MADPPjgg/3e//jjj/17cJ4gcFz5c0zryKIx6kgSF/xD0iI9BQoUKPAXwsLCmDZtGrNnzyYiIsLf6Qw7VFVVsXr1amJiYli5cqW/0xlSmEwmqqur0Wq1DptFgjgwuuyyy4bvwOeMM84gIiKCb775xv7eLbfcQkpKCo899hiTJk1iz549PP/888TGxg4qDqkzPkVFRSxZsmTQvJy1S/6vuq2CsNdPRGU1YvnDuwhLz3XNGSK4ug7BFFeuT0/4Ujju2g6pNgMYijbl8b2hzUNnfIxGI5GRkYP6cdY+nGZ8XF2HXjz++OM8/vjjA7adeOKJvPLKK5JmfNyNKwVGo5Fjjjlm0L2cXn31VS6//HKv5mQ0GhEEwSszPkG3j09lZSU//fQTX375pf29srIyXn75ZXJzc1mxYgUAq1evJjY2lp07d/Laa68N6CsyMnLAix8eHj7gH4ARI0Y4/cPgqt2Z736YshhO/DscfJKw/ffDkjMk7UDqS7hTZ7DElevTE74Ujru2Q6rNAIaiTXl8b2jTarWiUqkICQkhJCTE/v1gcNbe+1izKx/BAHdr+Otf/8oll1wyYNvIkSOZOXOmpOMbfHHtVCoVe/fuHfBoCYCpU6d6/Dt3xul9DdRXSdF60A183n33XaZMmcLZZ59tf89gMAD0u5ChoaFePdjM1b4H3twHAoATb4OsT6C9CmKehi0PeNe/h/B6nX6MK9enJ3wpHHdth1ybAQpFm/L4vtCmqz9wwT6gcRfu1jlhwoQBd1n2dVypPufOHfwYEnf4nnC89fc8qBRns9l499132b59u8NZH0uXLmXhwoXccMMNpKSkUFZWxrPPPsv+/fu54IILvBZfq9XKapeM8JFw5pPi9wkvQ1Oxd/17CK/X6ce4cn16wpfCcdd2yLUZoFC0KY/vTW32zkoMNivQC1ftwwX+qtMXceX69IRvNpu9dlBpUA18fvrpJ6qqqvrtVRAeHs7evXuZPHky5557LqtWrWLXrl28//77nHXWWV6L7+qzb5+sLViyFRadATYzfP+PgNjR2R9rKHwVV65PT/hSOO7a+kWbAQhFm/L43tBm7/EDvccLHLoW41C4ah8u8Fedvogr16cn/BEjRtg11feIC08QVB91nX766YOO+BYtWuT1nZoPRXp6utPdSl21ewSVCrY+AeUHxFf+/2DFBd6NIRE+qdNPceX69IQvheOurV+0GYBQtCmP7w1t9j5V29TURHh4ON3d3YwePfj6xK6urkHbbTYbJpOJnp6eoP9IzFmdwRZXrk9P+Hq9Hp1Ox6hRoxw+8fEEQflUl6/Q0dFBdHS0W6vCpcJsNrN3717OOusszxYs/vo4HHwCxs6Em1IgcoxX81Nw+EK2NhUoOAQmk4mKigrZazIEQaC7u5uRI0cG/VNdCuQjJCSE+fPnD7hNgpS/30E14+NvxMTEOP1vyFW7LJz4998WOleKC51P67//0FDBp3UOcVy5Pj3hS+G4a+tXbQYQFG3K43tLmxERESxatAiTyURqairr1q0b1I+zdrPZbI8T7INyV9chmOLK9ekJPzU1lRNOOMErM3/KjE8fuBoxWiwWp1Nsztq98l910Q/wySUQEgZ/TYDJ/lnP4Oo6BFNcuT494UvhuGvrd20GCBRtyuMr2vQdFG3K47viSJnxCe4PTYcYycnJstplY8mZsHgr2Cyw138LnX1e5xDGlevTE74Ujru2ftdmgEDRpjy+ok3fQdGmPL4361AGPhKwdOlSWe1ewdYnIGwEVByEvK98H28ADEmdQxRXrk9P+FI47toGhDYDAIo25fEVbfoOijbl8b1ZhzLwkYD6+npZ7V7B+HnixoYAP/4TjHrfxzwEQ1LnEMWV69MTvhSOu7YBoc0AgKJNeXxFm76Dok15fG/WoQx8JGDMGOdPUrlq9xpOuFUcAOnr4eBTQxOzD4asziGIK9enJ3wpHHdtA0abfoaiTXl8RZu+g6JNeXxv1qEMfCQgYLZeDx8JW38b8CS9Ao2FQxP3N/hrPw1fbb0+1HwpHHdtA0abfoaiTXl8RZu+g6JNeXxv1nF4KM5LaGtrk9XuVSw+A5acJS50HuIdnYe0Th/HlevTE74Ujru2AaVNP0LRpjy+ok3fQdGmPL4361AGPsDOnTtZvny5fV+B1tZW4uPjsVgsxMTEAOKeFRMnTiQ9PZ3a2lpKS0spKCigqamJ5ORkjEYjDQ0Ndtvu7m5SUlJoaGigqKiI0tJSDAYD6enpdHV1Ofg1m80kJCTQ3NxMXl4e5eXlVFdXk5mZiU6nc7C12WzExsbS3t5OwdyrsIVGQkUMVT+8QEtLy4B5d3Z2kpaW5pB3Y2MjycnJ9PT0ONj29PTY8y4sLKSkpIS6ujrS0tLo7OwkJiaGuXPnOuTd0tJCbm4uFRUVVFVVkZWVZc9bEASHrzqdjqysLKqqqqioqCA3N5eWlhYSEhLse3YcmnddXR0lJSVYLBYaGhpISUkZMO/k5GQaGxspKCigtLSU2tpah7x7bS0WC/Hx8bS0tNDT04NGo7Hn3d7eTmxsLDabzYGj0+nIzMykurqa8vJy8vLyaG5upqWlpV/eXV1dqNVq6uvrKS4upqioyJ53d3e3g06MRiPJyck0NTU55J2eno5er3ew7c27tbWVnJwcNBoNlZWVZGdnM378eOLi4vrl3dHRQUZGBhEREZSVldnzTkxMxGQy2W21Wi0GgwG1Wo1Wq7XnrdVqSU1NxWAwOPg1Go0kJSXR1NREfn4+ZWVl1NTUkJGRgV6vd7C1Wq3ExcXR1tZGdnY2lZWV9rzb2tqIi4vDarU6cPR6PRkZGdTU1FBWVkZ+fj5NTU0kJSVhNBodbA0GA6mpqWi1WqxWK8XFxWi1WtRqdb+8TSYTiYmJ9nutb94dHR397rW4uDja29vteWs0GnJycvr1Eb33hF6vd9pH9PV/aB9RXFxMfX09arWarq4uh9+9J33E+PHj7feaRqOx32vO+ojIyEi3+4jRo0cP2kf0zbulpcVpHzFnzpxB+wiNRkNbWxutra1u9RGFhYVe7yNyc3Nl9RG9eff9fbrqIw6915z1EQPl3bePCAsLIzs7m/b2dqd9RO+9Nlgf0XuvqdVqRo0aJauPmDhxouQ+YtasWU77iMJCCZ98CArs0Ol0AiDodLoB22NiYpzynbWbTCZhz549gslkkpVjPxx4UhB2jBWEZ5YIQk+Hd30PAlfXIZjiyvXpCV8Kx13bgNSmH6BoUx5f0abvoGhTHt8Vx9Xf775QNjDsA1cbINlsNqefMzpr99lGXOYeeOV4aKuADbfA6Y94z/cgcHUdgimuXJ+e8KVw3LUNSG36AYo25fEVbfoOijbl8V1xlA0MfYS4uDhZ7T5B+Ag462nx+6RXobHA5yH9UqeP4sr16QlfCsdd24DUph+gaFMeX9Gm76BoUx7fm3UoMz59ENCHlLrCp5dD4bcw7yTY/o14qrsCBW5gOP1XrWB4QdGmAnehzPj4CL0Lqzxt9ynOeEzc0VkTC7lf+DSUv+r0RVy5Pj3hS+G4axvQ2hxCKNqUx1e06Tso2pTH92YdysBHAlavXi2r3acYPxdOulP8/sf7oKfDZ6H8Vacv4sr16QlfCsdd24DW5hBC0aY8vqJN30HRpjy+N+tQBj4SUFFRIavd59hwC0w4Ajq1cPBJn4XxV52+iCvXpyd8KRx3bQNem0MERZvy+Io2fQdFm/L43qxDGfhIwMSJE2W1+xzhI2Brn4XODfk+CeOvOn0RV65PT/hSOO7aBrw2hwiKNuXxFW36Doo25fG9WYcy8JEAs9ksq31IsGgLLD0HBCvs9c2Ozv6q0xdx5fr0hC+F465tUGhzCKBoUx5f0abvoGhTHt+bdSgDHwno7u6W1T5kOPNxCBsJlXGQ87nX3furTl/ElevTE74Ujru2QaNNH0PRpjy+ok3fQdGmPL4361AGPhIwdepUWe1DhnFzYONvC533eX+hs7/q9EVcuT494UvhuGsbNNr0MRRtyuMr2vQdFG3K43uzDmXgIwHFxcWy2ocUG26BCQugswEOPOFV1/6q0xdx5fr0hC+F465tUGnTh1C0KY+vaNN3ULQpj+/NOpQNDPvA1QZIZrPZ6SZaztr9shFX6U/w4R9AFQp/iYWpK7zi1tV18BV8EVeuT0/4Ujju2gadNn0ERZvy+Io2fQdFm/L4rjjKBoY+QmJioqz2IcfCLbDsXHGh83d3em2hs7/q9EVcuT494UvhuGsbdNr0ERRtyuMr2vQdFG3K43uzDmXGpw+C+siKwdBeDS+vA0s3XPgGrL5k6GIrCAoMp/+qFQwvKNpU4C58NuNTUFDAjh07OOWUU1iwYAHTp09n1apVbN++nY8//hij0Sgr8UBHUG69Pm42bPqH+P2++6FHJ9ulsvW6PL5yLIDvoGhTHl/Rpu+gaFMef8iPrEhPT2fLli2sXbuWuLg4jjvuOP7+97/z8MMPc8UVVyAIAvfddx8zZszgySefDLoB0M6dO1m+fDnr1q0DoLW1lfj4eCwWi/1ix8TEsGTJEtLT06mtraW0tJSCggKamppITk7GaDRiMpnstt3d3aSkpNDQ0EBRURGlpaUYDAbS09Pp6upy8Gs2m0lISKC5uZm8vDzKy8uprq4mMzMTnU7nYGuz2YiNjaW9vZ2srCyqqqrQaDTk5ubS0tIyYN6dK6+mZ/Qs6Gqkfc/dFBQU0NjYSHJyMj09PQ62PT099rwLCwspKSmhrq6OtLQ0Ojs7iYmJ4eijj3bIu6WlhdzcXCoqKqiqqiIrK8uetyAIDl91Op0974qKCnveCQkJmM1mx7w7O0lLS6Ouro6SkhLGjh1LQ0MDKSkpA+adnJxMY2MjBQUFlJaWUltb65B3r63FYiE+Pp6WlhZGjx6NRqOx593e3k5sbCw2m82Bo9PpyMzMpLq6mvLycvLy8mhubsZms/XLu6urC7VaTX19PcXFxRQVFdnz7u7udtCJ0WgkOTmZpqYmh7zT09PR6/UOtr15t7a2kpOTg0ajobKykuzsbBYuXEhcXFy/vDs6OsjIyGD69OmUlZXZ805MTMRkMtlttVotBoMBtVqNVqu1563VaklNTcVgMDj4NRqNJCUl0dTURH5+PmVlZdTU1JCRkYFer3ewtVqtxMXF0dbWRnZ2NpWVlfa829raiIuLw2q1OnD0ej0ZGRnU1NRQVlZGfn4+TU1NJCUlYTQaHWwNBgOpqalotVrGjRtHcXExWq0WtVrdL2+TyURiYqL9Xuubd0dHR797LS4ujvb2dnveGo2GnJycfn1E7z2h1+ud9hF9/R/aRxQXF1NfX49araarq8vhd+9JH7Fw4UL3+4jf7rUZM2bY83bVR8yZM2fQPqJv3jabzWkfcdRRRw3aR2g0Gtra2mhtbXWrjygsLPR6H5Gbmyurj+jNu+/v01Ufcei95qyPGCjvvn3E1KlTyc7Opr293Wkf0XuvDdZH9N5rarWa2bNny+ojlixZIrmPWLNmjdM+orCwELchuIF58+YJO3fuFNra2pzaJSQkCJdcconw6KOPuuM24KDT6QRA0Ol0A7anpqY65TtrN5lMwp49ewSTySQrR49R8pMg7BgrCA+ME4T6HFmuXF0HX8EXceX69IQvheOubVBr04tQtCmPr2jTd1C0KY/viuPq73dfhLkzOCouLnbr89X169ezfv36YbsT58yZM2W1+xULT4Xl50P+/2DvnXDN96BSeeTKX3X6Iq5cn57wpXDctQ1qbXoRijbl8RVt+g6KNuXxvVmHWx919R30VFVVDfhRls1mo6qqqp/9cIJer5fV7nec8RiEj4KqRMj+r8du/FWnL+LK9ekJXwrHXdug16aXoGhTHl/Rpu+gaFMe35t1SH6cfd68eRx11FGUlZU5vN/U1MT8+fO9llggQnDxAJyrdr8jehZsukv8ft/90N3ukRt/1emLuHJ9esKXwnHXNui16SUo2pTHV7TpOyjalMf3Zh0e7eOzbNkyjj32WH7++WeH94e7gMeNGyerPSBw/E0wcRF0NcGBxz1y4a86fRFXrk9P+FI47toOC216AYo25fEVbfoOijbl8b1Zh+SBj0ql4pVXXuH+++/n7LPP5sUXX3RoG86orKyU1R4QCIuAs54Sv095A7Q5kl34q05fxJXr0xO+FI67tsNCm16Aok15fEWbvoOiTXl8b9YheeDTO6tz22238dVXX/Hvf/+b6667zv6o3nDGypUrZbUHDBacAssvAMEm7uhss0mi+6tOX8SV69MTvhSOu7bDRpsyoWhTHl/Rpu+gaFMe35t1yDqyYuvWrSQkJPDrr79yzjnneCungEVqaqqs9oDCGY9B+GioToLsTyVR/VWnL+LK9ekJXwrHXdthpU0ZULQpj69o03dQtCmP7806JB9ZcfLJJ/PVV185fN7W0tLCtm3b7Js6BSuG5ZEVzhD3H/hpB4yaBLekwchx/s5IgR8QkNpUoABFmwrch08PKf3111/7LTKaOHEiBw8eDOpBjzsYdluvH38jTFoMhmb49VG3acrW6/L4yrEAvoOiTXl8RZu+g6JNeXxv1uH2jE9HR4dbDr09UzKUcDViNBqNREZGDsp31h6w/7mUH4Bd54MqBK4/CNNXuaS4ug6+gi/iyvXpCV8Kx13bYalND6BoUx5f0abvoGhTHt8VxyczPuPGjWP8+PGDvnrbhzMyMzNltQckjtgMK7aJC533urfQ2V91+iKuXJ+e8KVw3LUdltr0AIo25fEVbfoOijbl8b1Zh1tHVoD4EVcvBEHgrLPO4q233jpsthsHOOKII2S1ByzOeBSKf4TqZMj6BNZe7tTcX3X6Iq5cn57wpXDctR222pQIRZvy+Io2fQdFm/L43qzD7RmfTZs22V+bN28mNDSU448/3uH9TZs2eS2xQERzc7Os9oDF2Bmw+W7x+/3/hu42p+b+qtMXceX69IQvheOu7bDVpkQo2pTHV7TpOyjalMf3Zh2yHmc/3ODqM+ag/gz6uL/CpCXiQudfnC909ledvogr16cnfCkcd22HtTYlQNGmPL6iTd9B0aY8vjfrUAY+wM6dO1m+fDnr1q0DoLW1lfj4eCwWi30leUxMDDabjfT0dGprayktLaWgoICmpiaSk5MxGo0UFhbabbu7u0lJSaGhoYGioiJKS0sxGAykp6fT1dXl4NdsNpOQkEBzczN5eXmUl5dTXV1NZmYmOp2uXw6xsbG0t7eTlZVFVVUVGo2G3NxcWlpaBsy7s7OTtLQ0h7wbGxtJTk6mp6dHtA2LIHvWlQAIqW/TkvsLhYWFlJSUUFdXR1paGp2dncTExDBy5EiHvFtaWsjNzaWiooKqqiqysrLseQuC4PBVp9PZ866oqLDnnZCQgNlsHjDvuro6SkpKaGhooKGhgZSUlN/z/s22p6eH5ORkGhsbKSgooLS0lNraWoe8e20tFgvx8fG0tLRQV1eHRqOx593e3m7flqEvR6fTkZmZSXV1NeXl5eTl5dHc3ExJSUm/vLu6ulCr1dTX11NcXExRUZE97+7ubgedGI1GkpOTaWpqcsg7PT0dvV7vYNubd2trKzk5OWg0GiorK8nOzsZisRAXF9cv746ODjIyMtDr9ZSVldnzTkxMxGQy2W21Wi0GgwG1Wo1Wq7XnrdVqSU1NxWAwOPg1Go0kJSXR1NREfn4+ZWVl1NTU2GP1tbVarcTFxdHW1kZ2djaVlZX2vNva2oiLi8NqtTpw9Ho9GRkZ1NTUUFZWRn5+Pk1NTSQlJWE0Gh1sDQYDqampaLVaGhsbKS4uRqvVolar++VtMplITEy032t98+7o6Oh3r8XFxdHe3m7PW6PRkJOT06+P6L0n9Hq90z6ir/9D+4ji4mLq6+tRq9V0dXU5/O496SMsFovkPqKzs3PwPqLPvdZXywP1EX3zLikpcdpHjBgxYtA+QqPR0NbWRmtrq1t9RGFhodf7iNzcXFl9RG/efX+frvqIQ+81Z33EQHn37SN0Oh3Z2dm0t7c77SN677XB+ojee633vpLTRxyagzt9RGRkpNM+ovf6ugXBQ4wZM0YoLy/3lB6Q0Ol0AiDodLoB29PS0pzynbWbTCZhz549gslkkpWjz7H7GkHYMVYQ3jxVEKzWAU1cXQdfwRdx5fr0hC+F467tYaFNN6BoUx5f0abvoGhTHt8Vx9Xf775we3Hztm3bHH7u6enhL3/5C6NHj3Z4/8svv3R/1BVkWLRokaz2oMDpj4gLnWtSIfMjOOrKfib+qtMXceX69IQvheOu7WGhTTegaFMeX9Gm76BoUx7fm3W4/VHX2LFjiY6Otr+uuOIKZsyY4fBedHS01xILRGRkZMhqDwqMnQGb7xG//2kHGFr7mfirTl/ElevTE74Ujru2h4U23YCiTXl8RZu+g6JNeXxv1iH5yIrhjMPuyIrBYDXDaydCUyEccy2c85y/M1LgQwSVNhUcVlC0qcBd+GQDwzlz5nDzzTezf/9+LBaL7CSDEYfN1uuh4XDWM+L36negznGkrWy9Lo+vHAvgOyjalMdXtOk7KNqUx/fLkRUHDx7k66+/5uuvv6apqYkzzjiD8847j7PPPrvf2V3BClcjRovFQljY4MuinLUH5X8un18LuZ/DzGPg2v0QIo6TXV0HX8EXceX69IQvheOu7WGnzUGgaFMeX9Gm76BoUx7fFccnMz6bNm3i2WefpaSkhPj4eNasWcNLL73EtGnTOOWUU/jPf/5DeXm5+1UEIZKTk2W1Bx1OfwQixkCtGjI/tL/trzp9EVeuT0/4Ujju2h522hwEijbl8RVt+g6KNuXxvVmHR/v4rFixgnvvvZekpCQqKiq49NJL+fnnnznyyCM58sgj+e6777yWYCBh2bJlstqDDmOnw+Z7xe/3/77Q2V91+iKuXJ+e8KVw3LU97LQ5CBRtyuMr2vQdFG3K43uzDtkbGE6fPp3rrruOb775hubmZh5++GG/nEA7FKitrZXVHpQ47gaYvAy6W+GXhwH/1emLuHJ9esKXwnHX9rDU5gBQtCmPr2jTd1C0KY/vzTpkffAnCAK//vor3d3dbNiwgfHjx3PhhRd6K7eAQ1RUlKz2oERoOJz9DLx3NqjfhbVXEhU1wS+p+OL6yvXpCV8Kx13bw1KbA8BfdSra9NxO0WbwxQ10bbqC2zM+7e3tbN++nZUrV3LdddfR0dHBSSedxJYtWzj33HNZtmwZ2dnZXkssEKFSqWS1By3mnQgrLwYE+O4OVPhnBwRfXF+5Pj3hS+G4a3vYavMQ+KtORZue2ynaDL64ga5NV3B74HPnnXeSmJjIn/70J3JycjjzzDOxWq0kJiaSnJzMsmXLuO+++7yW2KF44IEHUKlUDq+lS5fa27VaLVdeeSXTpk1j9OjRHHXUUXzxxRdezUGn08lqD2qc/jBEREFdOqFZH/klBV9cX7k+PeFL4bhre1hrsw/8VaeiTc/tFG0GX9xA16YruP1R1/fff8/HH3/Mpk2buPrqq5k9eza//PILxx13HABPPvkk5513ntcSGwgrVqzgp59+sv/c99G2q666ivb2dr7++msmTZrExx9/zMUXX4xarWbt2rVeiT9nzhxZ7UGNqGlw8j/hx3uZUfAmrNoE808a0hR8cX3l+vSEL4Xjru1hrc0+8FedijY9t1O0GXxxA12bruD2jE9DQwOLFy8GYObMmYwYMYLZs2c7JNXU1OS1xAZCWFgY06ZNs78mTZpkb0tISOCWW27h2GOP5YgjjuD+++9n3LhxpKWleS1+bm6urPagx7HXw9SVhPS0w/vnwAcX9tvc0JfwxfWV69MTvhSOu7aHvTZ/g7/qVLTpuZ2izeCLG+jadAW3NzAMCQlBq9UyZcoUQFxolJWVxRFHHAGIA6MZM2ZgtVq9llxfPPDAAzz99NNER0czYsQI1q9fz+OPP24fBZ5++ulERESwa9cuxo0bx2effca1115LVlYWCxcuHNCn0WjEaDTaf+7o6GD27Nk0NzcPuAGSzWYjJGTwsaKzdrPZzP79+znttNOCeyOurmZCYp4kJPMDVDZxB2/b0vOwbroXJvn2ED5X198fPj3hS+G4a6toU4QvNOKvuIo2FW0GatxA1GZHRweTJk1yawNDSU91vfXWW4wZMwYQd1F877337LMuer1eiivJOO6443jvvfdYsmQJ9fX1PPjgg5x00knk5uYSFRXFZ599xiWXXMLEiRMJCwtj1KhRfPXVV4MOegAef/xxHnzwwX7v79u3j1GjRvmkjv379/vE79DiZEYtXcHS+i+Z1ZZISOHXqAq/oWrCSRRNv4DuiEmuXSgIOAwPbSoYjlC0qcAVDAaD27Zuz/jMmzfPrVXVFRUVbgeXg/b2dubOnctzzz3Htddeyy233EJKSgqPPfYYkyZNYs+ePTz//PPExsaycuXKAX1InfGRg+H0n4sDGvMJPfAYISU/ACCERmA7+hpsG26D0coAKBgwbLWpIOihaFOBu5Ay44MQxDjmmGOEe+65RygtLRUAITc316H91FNPFW644Qa3/el0OgEQdDrdgO0HDx50ynfWbjKZhD179ggmk8ntfAIVA9ZZlSwI754tCDvGiq9HZwjCL48KQne7b+P62acnfCkcd20VbYrwhUb8FVfRpqLNQI0biNp09fe7L4b+A0cvobOzk7KyMqZPn26f4jr087/Q0FBsNpvXYq5Zs0ZW+3DBgHXOPha2fwNXfgXT14CpEw4+CS+shvgXwdztm7h+9ukJXwrHXVtFmyL8VaeiTc/tFG0GX9xA16YruDXwefHFF91++Qp33nknBw8eRKPRkJCQwIUXXkhoaCiXXnopS5cuZeHChdxwww2kpKRQVlbGs88+y/79+7ngggu8lkNZWZms9uGCQetUqWDBKXD9Abh4F0xaDN1tsP9f8OJR4s7PVrP348qAXJ+e8KVw3LVVtCnCX3Uq2vTcTtFm8MUNdG26gluLm59//nmHn5uamjAYDIwbNw4Q19uMGjWKKVOm8Le//c1ryfVFTU0Nl156KS0tLUyePJkTTzyRpKQkJk+eDMDevXu55557OPfcc+ns7GThwoW8//77nHXWWV7LoTeWp+3DBS7rVKlg+fmw5GzI/hQOPAG6avj275Dwkrgf0IptIHFVvy+ur1yfnvClcNy1VbQpwl91Ktr03E7RZvDFDXRtuoJbA5++C5Y//vhjXnnlFd5++22WLFkCQFFREddddx033HCD1xI7FJ9++qnT9kWLFnl9p+ZD0XchtCftwwVu1xkaBmuvgJUXgfodiHkGWsvgi2sh/j9wyr9h0WniQMmbcSVArk9P+FI47toq2hThrzoVbXpup2gz+OIGujZdQfIan3/961+89NJL9kEPwJIlS3j++ee5//77vZZYIKKnp0dW+3CB5DrDIuH4v8KtmXDyfRA5FrQ58PFF8O5WqEz0Tdwh8OkJXwrHXVtFmyL8VaeiTc/tFG0GX9xA16YrSB741NfXY7FY+r1vtVppaGjwSlKBiqlTp8pqHy7wuM7IKNh0F9yaBRtugbARUJUI754JH10E9c4PufXF9ZXr0xO+FI67too2RfirTkWbntsp2gy+uIGuTVeQPPA59dRTueGGG0hPT7e/l5aWxl//+le2bNnitcQCESUlJbLahwtk1zlqApz+CPwtA46+GlShULIPXj8JPv8ztAy8iM0X11euT0/4Ujju2iraFOGvOhVtem6naDP44ga6Nl3B7Q0Me9HU1MT27dv54Ycf7BtKWSwWzjjjDN577z37kRbBiI6ODqKjowfdAMlkMhERETEo31m72Wxm7969nHXWWUG/EZer6yAZLWXw66OQ+9saLVUoHHUlbLobxs7wXVwv+PSEL4Xjrq2iTRG+0Ii/4iraVLQZqHEDUZuu/n73heQZn8mTJ7N3716KiorYvXs3u3fvpqCggL179wb1oMcdJCUlyWofLvB6nRMXwB/fgRtiYdHpIFgh7T14cS3sux8Mrb6J6wWfnvClcNy1VbQpwl91Ktr03E7RZvDFDXRtuoLkGZ/hDCkjRqkYTv+5+ByVifDzg+L6H4CIKHFN0HE3wMhxfk1tOELRpoJAhaJNBe7C6zM+TzzxBN3d7u28m5yczHfffeeWbbAhJiZGVvtwgc/rnLservkeLtsNU1eCSQ8HHsP29ELYfQ2U7Adr/wX2nkBuLZ7wpXDctVW0KcJfdfoirqLN4QVFm/L43qzDrRmfq666iu+//56LLrqIc889l2OOOca+mZDFYiE/P5+4uDg+/PBD6urq2LVrFxs3bvRakkMFVyNGg8Hg9NR2Z+3D6T8XV9fBq7DZIP8rcQ+gxvzf3x8zFVZdDKsvg6nLPXYvtxZP+FI47toq2hQxpNr0cVxFm4o2AzVuIGrT6zM+u3bt4qeffsJsNnPZZZcxbdo0IiIiiIqKIjIykrVr1/LOO+9w1VVXUVhYGJSDHneQn58vq324YEjrDAmBI/8Af00g/6RX4dgbYOQE6GwQd4F+dT28vgmSX4euFsnu5dbiCV8Kx11bRZsi/FWnL+Iq2hxeULQpj+/NOtxe3Lx69WrefPNNWlpaSEtLY/fu3bz55pv8+OOPNDQ0oFar+ctf/sKIESO8ltxQYefOnSxfvpx169YB0NraSnx8PBaLxT69FhMTw/jx40lPT6e2tpbS0lIKCgpoamoiOTkZo9FIY2Oj3ba7u5uUlBQaGhooKiqitLQUg8FAeno6XV1dDn7NZjMJCQk0NzeTl5dHeXk51dXVZGZmotPpHGxtNhuxsbG0t7eTlZVFVVUVGo2G3NxcWlpaBsy7s7OTtLQ0h7wbGxtJTk6mp6fHwbanp8eed2FhISUlJdTV1ZGWlkZnZycxMTHMmjXLIe+WlhZyc3OpqKigqqqKrKwse96CIDh81el09rwrKirseSckJGA2mwfMu66ujpLSUnSj5tNw9B2kbvoQ07Z3aZ50PISEQX0mfH8XtmcWY3z/j1Tvf43SogJqa2sd8u71a7FYiI+Pp6WlhZ6eHjQajT3v9vZ2YmNjsdlsDhydTkdmZibV1dWUl5eTl5dHc3MzLS0t/fLu6upCrVZTX19PcXExRUVFNDQ0kJKSQnd3t4NOjEYjycnJNDU1UVBQQGlpKbW1taSnp6PX6x1se/NubW0lJycHjUZDZWUl2dnZREdHExcX1y/vjo4OMjIyCA8Pp6yszJ53YmIiJpPJbqvVajEYDKjVarRarT1vrVZLamoqBoPBwa/RaCQpKYmmpiby8/MpKyujpqaGjIwM9Hq9g63VaiUuLo62tjays7OprKy0593W1kZcXBxWq9WBo9frycjIoKamhrKyMvLz82lqaiIpKQmj0ehgazAYSE1NRavVYrFYKC4uRqvVolar++VtMplITEy032t98+7o6Oh3r8XFxdHe3m7PW6PRkJOT06+P6L0n9Hq90z6ir/9D+4ji4mLq6+tRq9V0dXU5/O496SOio6Ml9xERERFu9xEjR44ctI/om3dLS4vTPmLmzJmD9hEajYa2tjZaW1td9xElJRQWFtrvtYHyTk5OprGx0eFec9VH5ObmyuojevPu+/t01Ucceq856yMGyrtvHxEaGkp2djbt7e1O+4jee22wPqL3XlOr1YwYMUJWHzF+/HjJfcT06dOd9hGFhYW4Dclnww9juDrWvqioyCnfWbvJZBL27NkjmEwmWTkGAlxdhyGN29ksCEmvCcJrGwVhx9jfX0/OF4S9dwlCbYYg2GzSfMrNyYscd20VbYoIKG362aeizcCCok15fFccV3+/+8Kts7oUiBBcLIdy1T5c4K86B4w7eqL4tNdxN0BDPmR9DNmfiR+FJb8mvqYsh9WXimuCoqa59ik3Jy9y3LVVtCkioLTpZ5+KNgMLijbl8b1Zh+R9fA5nREdHy2ofLvBXnS7jTl0u7gh9Wz5c/rl4AnxopLgoev+/4Lll8OEfIfdLMPe451NuTjI57toq2hQRsNr0g09Fm4EFRZvy+N6sQxn4SEB1dbWs9uECf9XpdtzQMPHU94vehTuL4Zz/wKxjQbBB6X74/Bp4djF883cailKGJicPOe7aKtoUEfDaHEKfijYDC4o25fG9WYeygWEfKI+zu4egfSyzuRSyPoGsT6GjBgAhfCSqLQ/CuuvEJ8iGICflkWHfIWi16QOfijYDC4o25fGH/HH2vnj33XcxGAxSacMCarVaVvtwgb/qlB130kI49V/w9xy46muYewIqczd8fxe8d/agh6N6OycpHHdtFW2KCFpt+sCnos3AgqJNeXxv1iF5xmfq1Kl0d3dz0UUXce2117JhwwavJeNvKEdWHGaw2UD9NuzfAeYuCBsBp9wPx98IIaH+zm7IoGhTQaBC0aYCd+HTGZ/a2lref/99mpub2bx5M0uXLuXJJ59Eq9V6nHCwQNl6XcSw2Xo9JISYniVwYyIcsRksPeKhqG+fDo3u7QmhHAsQWBg22vSCT0WbgQVFm/L4Q35kxWBoaGjgww8/5P3336ewsJAzzzyTa6+9lnPPPZcQD9ZL+BuuRoxGo5HIyMhB+c7ah9N/Lq6uQzDFtfsUBEjfJQ58jB0QGgGb74ENt4qLpb2YkxSOu7aKNkUMS20OIV/Rpu+gaFMe3xXHpzM+fTF16lROPPFE1q9fT0hICDk5OWzfvp0FCxZw4MABOa4DEhkZGbLahwv8Vacv4tp9qlRw9Ha4MQkWnQ5WE/z8ELx1CmhzvZqTFI67too2RQxLbQ4hX9Gm76BoUx7fm3V4NPBpaGjgmWeeYcWKFWzevJmOjg6+/fZbKioqqK2t5eKLL2b79u1eSzJQsGDBAlntwwX+qtMXcfv5jJ4Jl30GF74OI8ZBfRa8sQl+fRwsJq/kJIXjrq2iTRHDWptDwFe06Tso2pTH92Ydkgc+5557LrNnz+a9997juuuuo7a2lk8++YQtW7YAMHr0aO64445huTdDU1OTrPbhAn/V6Yu4A/pUqWD1n+CmZFh6DtgscPAJePNkqMtwzfckpkxbRZsihr02fcxXtOk7KNqUx/dmHZKPrJgyZQoHDx5k/fr1g9pMnjyZiooKWYkFIlx9JumPz2/9AX/V6Yu4Tn1GTYNLPoS8L2HvP6AhF948FU64FTbdDeEjPMpJCsddW0WbIg4bbfqIr2jTd1C0KY/vzTokz/hs2rSJo446qt/7JpOJXbt2AaBSqZg7d6787AIMyg0s4rC6gVUqOPIPcFOKeASGYIW45+D1jVCdqvxxCTAcVtr0AV/Rpu+gaFMe368Dn2uuuQadTtfvfb1ezzXXXOOVpAIVypStiMNyynb0JPEIjEs+hNFToLkI3jmdkfv+4faj75JjSrBVtCnisNSmF/mKNn0HRZvy+N6sQ/LARxAEVCpVv/dramqC9rC5nTt3snz5ctatWwdAa2sr8fHxWCwW+94BMTExTJs2jfT0dGprayktLaWgoICmpiaSk5MxGo00Nzfbbbu7u0lJSaGhoYGioiJKS0sxGAykp6fT1dXl4NdsNpOQkEBzczN5eXmUl5dTXV1NZmYmOp3OwdZmsxEbG0t7eztZWVlUVVWh0WjIzc2lpaVlwLw7OztJS0tzyLuxsZHk5GR6enocbHt6eux5FxYWUlJSQl1dHWlpaXR2dhITE8PChQsd8m5paSE3N5eKigqqqqrIysqy5y0IgsNXnU5nz7uiosKed0JCAmazecC86+rqKCkpAcSF9SkpKQPmnZycTGNjIwUFBZSWllJbW+uQd6+txWIhPj6elpYWLBYLGo3Gnnd7ezuxsbHYbDYHjk6nIzMzk+oxa9Cc8xntc88EwcbEqr3wynG0vnQKlB8g5uBBurq6UKvV1NfXU1xcTFFRkT3v7u5uB50YjUaSk5NpampyyDs9PR29Xu9g25t3a2srOTk5aDQaKisryc7OZvLkycTFxfXLu6Ojg4yMDEaNGkVZWRl5eXk0NzeTmJiIyWSy22q1WgwGA2q1Gq1Wa89bq9WSmpqKwWBw8Gs0GklKSqKpqYn8/HzKysqoqakhIyMDvV7vYGu1WomLi6OtrY3s7GwqKyvtebe1tREXF4fVanXg6PV6MjIyqKmpoaysjPz8fJqamkhKSsJoNDrYGgwGUlNT0Wq1qFQqiouL0Wq1qNXqfnmbTCYSExPt91rfvDs6Ovrda3FxcbS3t9vz1mg05OTk9Osjeu8JvV7vtI/o6//QPqK4uJj6+nrUajVdXV0Ov3tP+ogpU6ZI7iPGjBnjdh8RHR09aB/RN+/29nanfcSCBQsG7SM0Gg1tbW20tra61UcUFhZ6vY/Izc2V1kdUV1NeXm6/13rz7vv7dNVHHHqvOesjBsq7bx8xYsQIsrOzaW9vd9pH9N5rg/URvfeaWq1m7NixsvqIadOmSe4j5s+f77SPKCyU8A+o4CbWrFkjrF27VggJCRFWrlwprF271v5atWqVEBUVJVx00UXuugtI6HQ6ARB0Ot2A7QcPHnTKd9ZuMpmEPXv2CCaTSVaOgQBX1yGY4sryWZkkNL18hiDsiBaEHWPF16snCELmJ4JgNnolpru2ijZFKNqUx1e06Tso2pTHd8Vx9fe7L9zewPDBBx+0f73jjjsYM2aMvS0iIoJ58+bxhz/8gYiICPdHXQEG5cgKBR6hpQySX4OMD8H82zl2UdPh2OvhmGtg5Hj/5ucCijYVBCoUbSpwFz7ZwHDHjh3s2LGDd999l4cfftj+844dO7j33nu59NJLg3rQ4w6UrddFKFuvH8KfuADOehpuy4NT/w1jpoG+Hn5+EJ5bAXvvgtYKR46X81O0KULRpjy+ok3fQdGmPH7AHFkx3OBqxGi1WgkNHfzwSmftw+k/F1fXIZjiyvU5IN9igtwvIPFl8RF4AFUILD0b1t+CdeYxbsd0Nz9FmyIUbcrjS+Eo2pQGRZvy+K44Xp/xmTBhgn1h1vjx45kwYcKgr+GMxMREWe3DBf6q0xdx5fockB8WAWsuhb/EwZV7YOEWEGxQ8A28czrGZ1fBnptA/Y64M7TVIjs/RZsiFG3K40vhKNqUBkWb8vjerMOtDQyff/55oqKi7N8P9FTX4YAVK1bIah8u8Fedvogr16dTvkoFC04WX40FkLgTsv/LKEMNZH4ovgDCRsL01TDrGJh5tPgaNwdUKrfzU7QpQtGmPL4UjqJNaVC0KY/vzTrcGvj0PXfr6quv9lrwYEN1dTXjxw++UNVV+3CBv+r0RVy5Pt3mT1kG578Mpz2EJuZT5oW3QK0aajPAqIPqJPHVi9GTYebRGENnwuotMG0lRM8SB1NS8jD3QFMBbXlqxp/6ZwgZ+qn2oYSiTXl8KRx3bZV+U4SiTXl8b9Yh+ciK9PR0wsPDWblyJQD/+9//ePfdd1m+fDkPPPDAsF7g7GqfomDdx0gq/FWnL+LK9SmZP2oCqmVnQ+/O5jYbtJRCbZo4EKpRi+uCupqg+AemARS8LdqOGCcOgKYeKX6ddiRMXgphkWIeXS3QkAPaPq+mIhCsHAHQniAevhp2+N6jwRR3yLUpkeOurdJvilC0KY/vzTokD3xuuOEG7rnnHlauXEl5eTmXXHIJ27ZtY/fu3RgMBv7zn/94LTkFCoY9QkJg8mLxteZS8T1zD2izoUZNZ2k8Yzo10FQIPe2giRVfdn4YTFzEzK5WMDQMHGPkBARjB6q8L8HUCRfvgvCRvq5MgQIFCgISkgc+xcXFrFmzBoDdu3ezadMmPv74Y+Lj4/nTn/40rAc+Ax3VIaV9uMBfdfoirlyfnvBdcsJHwOxjYfaxlI8+kVWrVoHFKM7eaHPEGaHeGZ2edmgq+P1GnnDEb7NBK2HaKvFr1HQq9r/FESn/gpJ98OEf4dJPYIR396oKBCjalMeXwnHXVuk3RSjalMf3Zh2SBz6CIGCz2QD46aefOOeccwCYPXu2/cmv4YrZs2fLah8u8Fedvogr16cnfCkcu21YJExfJb56IQigq4HGAjpMMHbh8YMOZsav+yMsWQEfXwKVcbDrfLjiCxg1vJ7EVLQpj++RNmXaKf1m8MUNdG26guSzuo455hgeeeQRPvjgAw4ePMjZZ58NQEVFBVOnTvVaYoGIvLw8We3DBf6q0xdx5fr0hC+F49RWpYJxs2Hx6WS3j3I6g5OXlwdzN8D2b2DkBKhLh3fPAr1WSuoBD0Wb8vhe06YEO6XfDL64ga5NV5C8gWF2djaXX345VVVV3H777ezYsQOAW265hZaWFj7++GOvJTfUUDYwdA/KRlzy+H7fJK6xED64QNxdevx8zJd9wd6EXEWbARb3sNTmIVD6zcCMG4ja9MmRFb1YtWoVOTk56HQ6+6AH4Omnn+b999+X6i6oEB8fL6t9uMBfdfoirlyfnvClcNy1laTNKUvhzz/A+HnQVkHYrrMZ01Prdk6BDEWb8vh+1+YwhqJNeXxv1uHxkRUmk4nGxkb7ep9ezJkzxyuJ+QPKIaUKDit01IszP02F2AiF+ScSsuxcWLJV3DNIgQI/Q+k3FbgLn874FBcXc9JJJzFy5Ejmzp3L/PnzmT9/PvPmzWP+/PkeJ+1P7Ny5k+XLl7Nu3ToAWltbiY+Px2Kx2A9Gi4mJYf/+/aSnp1NbW0tpaSkFBQU0NTWRnJyM0Whk165ddtvu7m5SUlJoaGigqKiI0tJSDAYD6enpdHV1Ofg1m80kJCTQ3NxMXl4e5eXlVFdXk5mZiU6nc7C12WzExsbS3t5OVlYWVVVVaDQacnNzaWlpGTDvzs5O0tLSHPJubGwkOTmZnp4eB9uenh573oWFhZSUlFBXV0daWhqdnZ3ExMTYX715t7S0kJubS0VFBVVVVWRlZdnzFgTB4atOp7PnXVFRYc87ISEBs9k8YN51dXWUlJTwxRdf0NDQQEpKyoB5Jycn09jYSEFBAaWlpdTW1jrk3WtrsViIj4+npaWF3bt3o9Fo7Hm3t7cTGxuLzWZz4Oh0OjIzM6murqa8vJy8vDyam5v58MMP++Xd1dWFWq2mvr6e4uJiioqK7Hl3d3c76MRoNJKcnExTU5ND3unp6ej1egfb3rxbW1vJyclBo9FQWVlJdnY2P/74I3Fxcf3y7ujoICMjg2+++YaysjJ73omJiZhGTCRhyf3Y5mwgBCshFQdh753w/Ap6XjiO5i/+QXPeAVJTUjAYDA5+jUYjSUlJNDU1kZ+fT1lZGTU1NWRkZKDX6x1srVYrcXFxtLW1kZ2dTWVlpT3vtrY24uLisFqtDhy9Xk9GRgY1NTWUlZWRn59PU1MTSUlJGI1GB1uDwUBqaiparZYvv/yS4uJitFotarW6X94mk4nExET7vdY3746Ojn73WlxcHO3t7fa8NRoNOTk5/fqI3pder3faR/T1f2gfUVxcTH19PWq1mq6uLoffvSd9xL59+yT3Ed9++63bfcTevXsH7SP65v3hhx867SMOHjw4aB+h0Whoa2ujtbXVrT6isLDQ631Ebm6urD6iN+++v09XfcSh95qzPmKgvPv2EV9//TXZ2dm0t7c77SN67zWHPsJk6nevqdVqvvvuO3veWq2W1NRUSX3E/v37JfcRBw4ccNpHFBYW4i4kz/iccMIJhIWFcc899zB9+vR+x1esXr1airuAgqsRo16vtx/dMRCctQ+n/1xcXYdgiivXpyd8KRx3beVq8+CX73DyDAOhxT9AdTLQp1sYES1+LDZurvh1/G9fJy+D6Jlu1TFUULQpjx+I2lT6zcCLG4ja9OmMT2ZmJq+//jpbt25lzZo1rF692uE1nFFaWiqrfbjAX3X6Iq5cn57wpXDctZWrza4R07AdfzNc+yPcWQLnvQxLzoKwEdCjEw9TLfgaEl6E7+6AD/8Az6+AvXeBUe92Pb6Gok15/EDU5nCBok15fG/WIXkfn+XLlw/7/XoGw+TJk2W1Dxf4q05fxJXr0xO+FI67tl7V5pjJcNSV4svcA20V0KaBtkrxa3sltFZAUwGkvC6eOn/W07DsHPdj+AiKNuXxA16bQQxFm/L43qxD8sDnySef5K677uKxxx5j5cqV/aYfvb0oOJBgNBpltQ8X+KtOX8SV69MTvhSOu7Y+02b4CPGA1SnL+reV/QLf3i4OjP57OSw9RxwAjZ3hWSwvQNGmPH5QaTPIoGhTHt+bdUge+GzZsgWAU0891eF9QRBQqVRYrVbvZBaAUG5gEcoNLI8/bP64LDgFbkyEg0+JH4EVfgvlB+GoqyBilKNtSJi4RmjiQpi4wGc7RivalMcfNtoMQCjalMf368Dn119/9VrwQMUDX+cSOWpMv/e7u3sYWZg9KM9Zu81mo6oqhIT/5RESInlpVUAgMiyUi46ZxRRlylYWf1h9nBA+ErbsgJV/hG9uhZpUSNrpmjdygjiLtP5mWHqW19JRPk6Qxx9W2gwwKNqUx/frR12bNm3yWvBAxedptYREjnJtKBkhJDYG90ZxHyRVcvaCSJ66ajMjwod2F9KysjKv38RyfXrCl8Jx19aVnS+unQOmroA/74PsT6Eus3+7pVtcF9RSBvo66G6FynjxtfhMOPMJmCB/Owyf1zmEcRVtDi8o2pTH92YdHm1gGBsby+uvv055eTm7d+9m5syZfPDBB8yfP58TTzzRK4n5A72Pwz31dTojR/ef8bFYrISFDf7H3lm71WqjuLiIxYuXEBoanDM+2TU69uU3ADB34igev3AlGxZOGrL4RqORyMjIgPLpCV8Kx11bV3bO2of8kWFjJ7SWQ+4XkLgTbGYIjYSTbocTbhVnkTx17QON+Cuuos3h9Ti7ok15fFccSRsQCxLx+eefCyNHjhT+7//+T4iMjBTKysoEQRCEl156Sdi6datUdwEFnU4nAIJOpxuw/eDBg075ztpNJpOwZ88ewWQyycrR39iXpxXW7PhOmHv3t8Lcu78V7vwsU2jrMg5JbFfX3x8+PeFL4bhrG7TabCwShPfOFYQdY8XXC2sEoSbNY3e+0Ii/4iraHD79piAo2pTLd8Vx9fe7LyTP+Kxdu5bbbruNq666iqioKLKysjjiiCPIyMhg69ataLXBe9qzcmSFe9D3mHnqhyI+TK5EEGDSmAj+fe4Kzl3Vf0NLBYEPv2tTECDvK/jxPvFjsJBwOONROPZ68QT6wdCmgZzdsOoSGBe8R+UoGBx+16aCoIFPNzAsKipi48aN/d6Pjo6mvb1dqrugQu922Z62DxdkpCTy8AVH8vlf1rNwyhiaO0387ZMMrn1fTW17t8/i+uL6yvXpCV8Kx13boNamSgVHbhOfEFt6jvjR1/d3wWdXQnd7f3tzNxx4AnYeB788Ap9eDjbxaVJ/1alo03O7gNamF6FoUx7fm3VIHvhMmzZtwB0U4+LiOOKII7yS1EB49dVXWbVqFWPHjmXs2LGsX7+e77//3t6u1Wq58sormTZtGqNHj+aoo47iiy++8GoOxxxzjKz24YLeOo+eO4Hv/nYif9+yiPBQFb8UNnL6cwd5L74Cq82js2/dihtIPj3hS+G4azsstDlyHFzyIZz5pDjrU/ANvH4S/HAvxL8IOZ9D1n/hlePhwONg6QFUoM0G9TuA/+pUtOm5XVBo0wtQtCmP7806JA98rrvuOm699VaSk5NRqVTU1dXx0Ucfceedd/LXv/7Va4kdilmzZvHEE0+QlpaGWq3mlFNO4fzzzycvLw+Aq666iqKiIr7++mtycnLYtm0bF198MRkZGV7LoTeWp+3DBX3rjAwL5e9bFrP3bydx9NzxdJmsPPBNPn94NYEirXePMvDF9ZXr0xO+FI67tsNGmyoVHP8X8eiMcXOgvQqSXoH9/4IvroWvrhc/4oqaAX98F7Y+JfJ+eRi6mn+vs6kYXj0RvvqrfTbIl1C06bld0GhTJvxV5+GoTVeQ/Dj7Pffcg81m49RTT8VgMLBx40YiIyO58847ueWWW7yW2KE499xzHX5+9NFHefXVV0lKSmLFihUkJCTw6quvcuyxxwJw//338/zzz5OWlsbatWu9ksPs2bNltQ8XDFTnoqlR7L5hPR+lVPHk94VkVrdz9oux/HXzAm46eaFXHn33xfWV69MTvhSOu7bDTpszj4YbYsU1PO2V0FEP+nowtMLiM2DjPyByDFgtkL4LGnLg5weZfex94iPz758LnVrx/ehZcMp9Pk1X0abndkGnTQ/hrzoPR226guSBj0ql4r777uMf//gHpaWldHZ2snz5csaM6f/4t69gtVrZvXs3XV1drF+/HoANGzbw3//+l7PPPptx48bx2Wef0dPTw+bNmwf1YzQaHXaD7OjoAMQFdWazuZ99a2srEydOHNSfs/ZefwP5DTY4q/NPR89g08IJPPRtAT8VNvHSL6V8l13Hw+cv59h58nbrdXX9/eHTE74Ujru2w1KbYaNh7dWDt/+Wr+qMJwjbdTZC+gdYIhch5L2GqlOLMGYaqk4txDyFZdpqhEVn+CxVRZue2wWlNj2ALzTir7iBqE0pGvFoHx9BEGhpaUGlUg3pLzInJ4f169fT09PDmDFj+PjjjznrLHHX1/b2di655BL27dtHWFgYo0aNYvfu3Zx++umD+nvggQd48MEH+73/8ccfM2qULzYwPHwgCJDVquKLihA6zOKTORum2Dh3ro1RkofbChQ4x9rK15nTGm//WR85nfhF/2Sx9n8c0fwTptBRHFzyEIbIKX7MUoECBb6CwWDgsssuc+upLkkDH61Wy1133cXXX3+NXi+u3xg7diwXXnghjz/+OFOnTpWXuQuYTCaqqqrQ6XR8/vnnvPXWWxw8eJDly5dzyy23kJKSwmOPPcakSZPYs2cPzz//PLGxsaxcuXJAfwPN+MyePZvm5uYBL1xDQ4PTGp21m81m9u/fz2mnnRb0j2W6ug590dFt5ql9xfxXLe5YPSUqkn+fvZQzVkjXipS4Q+XTE74Ujru2h702OxsJe+04VEY9wvj5WK74GsZOB6uJ0A/OI6RWjTDlSCzX/ABhI7weXtGm53bDXpu/wRca8VfcQNRmR0cHkyZNcmvg4/b/3h0dHWzYsIHOzk6uueYali5diiAI5Ofn88knnxAXF0d6erpPP/KKiIhg4cKFABx99NGkpqbywgsvcNddd/Hyyy+Tm5vLihUrAFi9ejWxsbHs3LmT1157bUB/kZGRA+4EGR4ePuBNptVqmTVr1qD5uWp35juY4E6dvZgYHs6Tf1zDhUfN5p9f5lDe3MXNn2Zx+vKpPHT+kUyLdv+PkJS4Q+XTE74Ujru2h702x8+Ei3fRdOANJv/hacLH/bYeIDwcLt4Fr29E1ZhLeM6ncOx1Xg+vaNNzu2Gvzd/gC434K24galOKPtx+quuFF14gNDSUvLw8nn/+eW644Qb+8pe/8OKLL5KXl4cgCLz44otuB/YGbDYbRqMRg8EA0O/wz9DQUGw2m9fiLV++XFb7cIEndR5/xET23noSN5+8kLAQFfvyGzjtuYN8mFSJzc1H331xfeX69IQvheOuraJNYMEpjL70HRh3yCLI6Jmw6S7x+6RXfPKUl6JNz+0OC23ivzoPR226gtsDn++++45//vOfAx4SNmXKFO69916++eYbryV2KO69915iYmLQaDTk5ORw7733cuDAAS6//HKWLl3KwoULueGGG0hJSaGsrIxnn32W/fv3c8EFF3gtB7VaLat9uMDTOkeEh3LnGUv49m8nsnr2OPRGC/fvyeWSNxIpbXT96Lsvrq9cn57wpXDctVW0KWLQOtdcDiPGiWeEFe0durh+9KloM7DgrzoPR226gttrfCZMmEBiYiJLliwZsL2wsJANGzbQ2trqteT64tprr+Xnn3+mvr6e6OhoVq1axd13381pp50GQElJCffccw9xcXF0dnaycOFC7rzzTq688kq3YyhHVgwdrDaBXYkanv6xCIPJSkRoCDedvJC/bl5ARFhwHuIarDhstPnzQxD7LMw+XtwnSEHA47DRpgLZ8MmRFR0dHYwbN27Q9nHjxtkfB/cF3n77bTQaDUajkcbGRn766Sf7oAdg0aJFfPHFFzQ0NNDV1UVWVpakQY87ULZeF+GNOkNDVFxzwnz2376Jk5dMxmS18fxPxZzzUixplW0+i+ttn8qxAIEFp3Ueez2ERkB1ElSnDF1cP/lUtBlYUI6skMf3Zh1uz/iEhoai1WoH/KgLxBXXM2bMwGr1/S6pvoKrEaPJZCIiImJQvrP24fSfi6vrIBWCIPBNdj0Pfp1HS5cJlQquPH4u/zhjCVEjfr9W3o7rDZ+e8KVw3LVVtCnC5fX6302Q8SEsOw8u+WDo4vrBp6LNwIIvNOKvuIGoTZ/M+AiCwOLFi5kwYcKAr6VLl7pfQZAiLS1NVvtwgbfrVKlUnLd6Bj/dvok/Hj0LQYBdiZWc/nwMP+U3+CyuN3x6wpfCcddW0aYIl3Wuv1n8WvituN5nqOL6waeizcCCv+o8HLXpCm4/zv7uu+96LWiwYtGiRbLahwt8Vef40RE8c9FqLlgzk39+lUNVq4H/26Xm7JXT2XHecp/ElevTE74Ujru2ijZFuKxzyjJYeBqU7od9/xIPRVWpfB/XDz4VbQYW/FXn4ahNV3B74LN9+3avBQ1WNDQ0MGnSJI/bhwt8XeeJiybx49838p+fi3krtoLvcuqJLWni6jXR3Hb+RFRe+EPVC7m1eMKXwnHXVtGmCLfq3PIAlB8QZ31ydsOqi4cm7hD7VLQZWPBXnYejNl1BeXxGAkaMcL7Znqv24YKhqHNkRCj3bl3G/246gZUzo+nosfBiUguXvplERXOX1+LIrcUTvhSOu7aKNkW4Vee0I2HT3eL3e+8UD0B1hp4OsBidmvji+iraHF7wV52HozZdQTk1SQIG2uVZSvtwwVDWeeTMaL66cQPvxmt4Zl8hSeWtnPGfGG49dRHXbzyC8FB5Y3e5tXjCl8Jx11bRpgi36zzxNnHGpz4TvvkbnHSnOAtUnwWjxkP0bEAFZb9ATYp4zMXZz8KaywZ0NyLUBj/eJ54ef/azECH/rD9Fm8ML/qrTF3EDXZuuoMz4SEBTU5Os9uGCoa4zLDSE6zYewX/OmMxJiyZhsth4+scizn0pjqzqdlm+5dbiCV8Kx11bRZsi3K4zNAwufE18vL1kH7xzOhx4DIq+E5/6OvC4+HN1Egg2MBtgz1/Fp8JS34bd18D750L8C6CJZ8KXl0Diy5D1MXx/19DW4kW+ok3fwV91+iJuoGvTFZQZHwlYsGCBrPbhAn/VecLqJZx5YhRfZdTy8Lf5FGr1XPhKPFdvmM8dpy9mdKR0OcutxRO+FI67too2RUiqc8oyOO0h+OEeGDke5m+C2ceBqRN0NeJgZ+4GWHAKZP1XHAxlfCi+elEh7i0yEmDkBOhph4wPYN6JsPpPQ1eLl/iKNn0Hf9Xpi7iBrk1XUGZ8JCAzM1NW+3CBv+rMzMxEpVKx7ahZ/HT7Ji5YMwObAO/EV3D68zEcKGr0yKfcnHzJcddW0aYIyXUe/1e4swT+UQ4Xvw/rbxTP9TrvRfjDW3DMn2H8PNh8N1z1P5h1LMzfCJv/CVufhnkngSqE9nEr4K8JsOke0e+3t4kffeV+CebuoanFC3xFm76DP/vNQPPpa226glsbGN5+++1uO3zuuedkJeRPKEdWBB8OFDVy31e51LaLf1zOXzODf5+znIljDo91A96Aok2ZMHVB+CjxsXibFT78A5T/+nv79NVw5R4YNcFvKQYrFG0qcBde38AwIyPDrVewjtx37tzJ8uXLWbduHQCtra3Ex8djsVjs22THxMSwf/9+0tPTqa2tpbS0lIKCApqamkhOTsZoNLJr1y67bXd3NykpKTQ0NFBUVERpaSkGg4H09HS6uroc/JrNZhISEmhubiYvL4/y8nKqq6vJzMxEp9M52NpsNmJjY2lvbycrK4uqqio0Gg25ubm0tLQMmHdnZydpaWkOeTc2NpKcnExPT4+DbU9Pjz3vwsJCSkpKqKurIy0tjc7OTmJiYuyv3rxbWlrIzc2loqKCqqoqsrKy7HkLguDwVafT2fOuqKiw552QkIDZbB4w77q6OkpKSuxHkqSkpNjz3rxkCjuODWX78bNRAf/LrOPkp3/hle/TqKmpcci716/FYiE+Pp6WlhZ2796NRqOx593e3k5sbCw2m82Bo9PpyMzMpLq6mvLycvLy8mhububDDz/sl3dXVxdqtZr6+nqKi4spKiqy593d3e2gE6PRSHJyMk1NTRQUFFBaWkptbS3p6eno9XoH2968W1tbycnJQaPRUFlZSXZ2Nj/++CNxcXH98u7o6CAjI4NvvvmGsrIye96JiYmYTCa7rVarxWAwoFar0Wq19ry1Wi2pqakYDAYHv0ajkaSkJJqamsjPz6esrIyamhoyMjLQ6/UOtlarlbi4ONra2sjOzqaystKed1tbG3FxcVitVgeOXq8nIyODmpoaysrKyM/Pp6mpiaSkJIxGo4OtwWAgNTUVrVbLl19+SXFxMVqtFrVa3S9vk8lEYmKi/V7rm3dHR0e/ey0uLo729nZ73r2HJDv0EUlpxMTGinl3GchYdg+tmx6nffFFWCKioT6Lrle3YGyrd/B/aB9RXFxMfX09arWarq4uh9+9J33Evn37JPcR3377rdt9xN69ewftI/rm/eGHHzrtIw4ePDhoH6HRaGhra6O1tdWtPqKwsLBfH9E37+TkZBobGx3uNVd9RG5urqw+ojfvvr9PV33Eofeasz5ioLz79hFff/012dnZtLe3O+0jeu+1wfqI3ntNrVbz3Xffyeoj9u/fL7mPOHDggNM+orCwEHfh9pEVhwNcjRhtNhshIYOPFZ21D6f/XFxdB3/Fzapu554vcyioF8+MO2nRJB69YCVzJg7+hI3cWjzhS+G4a6toU0TAabOxQFwE3dUkfkz25x8gJFR8+ksVAiPHSfcpNycvcRRtSkPAadOPPn2hTZ8cWaEAEhISZLUPF/irTldxV88ex9c3n8BdZy4hMiyE2JJmTv/PQd6IKcNitXnkU25Ocjnu2iraFBFw2pyyDK7+DiLHio/Fp7wJNWnw/ArYeSx0Dr4uTdHm8ELAadOPPn2tTVfw6KkutVrNZ599RlVVFSaTyaHtyy+/9EpigYgjjzxSVvtwgb/qdCdueGgIN25eyNYjp/PPL3NILG/hsb2FfJ1VxxPbVnHkzGjJPuXmJIfjrq2iTREBqc3JS8Tdor+7HX5+CCJGi0+MmQ3w3R2DHpaqaHN4ISC16SefvtamK0ie8fn000/ZsGEDBQUFfPXVV5jNZvLy8vjll1+Ijo527SCIUVVVJat9uMBfdUqJO3/SaD6+7jie+sMqxo4II7e2g/N3xvP49wV0m6we+ZSbkyccd20VbYoIWG0efQ3MPh7MXdDVCBMXQkgYFHwNeXs88yk3J5kcRZvSELDa9INPX2vTFSQPfB577DGef/55vvnmGyIiInjhhRcoLCzk4osvZs6cOV5LLBDhamA33Ad+vfBXnVLjqlQqLl43m5/u2MTZq6ZjtQm8frCcM/4TQ1xJs0c+5eYkleOuraJNEQGrzZAQOPcF8emvMdPEp7xOukNs++ZWqMuQ7lNuTjI5ijalIWC16QefvtamK0ge+JSVlXH22WcDEBERQVdXFyqVittuu4033njDa4kFIlytAz9c1on7q05P406JGsHOy47i7e3HMD16BFWtBq54O5m/f5pBq8E85DlJ4bhrq2hTREBrc8pSuCUNbkyEcbPFYzJmHyduevj++VCjdu2zow5in4PWCu/kJIOjaFMaAlqbQ+zT19p0BckDn/Hjx6PX6wGYOXMmubm5ALS3t2MwGLyWWCCit25P24cL/FWn3LinLpvK/ts3cfWGeahUsCezjis+KeaTlCpsNs9uKk9yksJx11bRpoiA1+bYGb/v5xMWAZd/DnPWg1EHu68G2++L8O0+k9+Az6+FrE/hrS3w84Pw+iYo+sFpKFNNJmR+DBL+YCja9B0CXptD6NPX/aYrSB74bNy4kf379wNw0UUXceutt3Lddddx6aWXcuqpp3otsUDEzJkzZbUPF/irTm/EHRMZxgPnreB/N53AkTPH0mUSuPfLHC5+PZEirfQby5OcpHDctVW0KSLotDlirDj4CR8NumrQZjv6bNPAD3dD7ufw1Q3QUSueL2bUwX8vh9bygf0KAgvSHhTPFyva65M6FG1KQ9Bp04c+fd1vuoLkgc/LL7/Mn/4knkFz3333cfvtt9PQ0MAf/vAH3n77ba8lFogoKCiQ1T5c4K86vRl31axx7LnxBK5YMYLREaGoK9s4+8VYnvyh0GHxsy9yksJx11bRpoig1GbkGDhis/h9yX5Hn4mviIekjpsDEVHi7NDfc8SPyGwWKPh2YJ/NJYS2/TYoyvvK7VQGrSPzk34zTIo2pSEotekjn77uN11B0gaGFouFjz/+mDPOOIOpU6d6LYlAgasNkCwWC2Fhg+8A4Kx9OG3E5eo6BFNci8VCY6eZB77OY19+AwCzJ4zk4fOPZPOSKT7JSQrHXVtFmyKCVptp74mLnGcdC/8nDn4sHY2EvbRafOz9yj3iIaoqlfhKfl08BX7uiXDNd/39xb8I+/8lfh8RBf8ohfARntXRkA+vrhe//1czhIYPbuuuTzfbFW0GZly5Pn3Rb/psA8OwsDD+8pe/0NPTI4U2bKBsxCViuG3ENWPcSN646hjeuPJoZkSPoLq1m6vfTeXmj9Np7HCudWWTuMBC0Gpz4Wni15pU6GoRv/3qAXHQM22lOCMUEiIOegAWnyF+rUqE7rbf/VQlQeyzjrM8Jj2U/dI/Zkc95H8NVovzOuozf/9eV+3cdgAo2hQRtNr0gU9/b2Ao+ciKzZs3c9ttt3H++ed7LYlAgXJIqYIuo4Xn9xfzTnwFNgGiIsO4a+tSLj92DiEhKn+n5xMo2gwQvHoCNOTCtjdh4RZ4YY24lucPb8PKP/a333k8NBWI7fM3wf5/Q9bHjjZLzoai72DNFXDBTse2jy6Ckn0w8xi4fPfgh6h+dwekviV+f/kXsGiL7FLdhaJNBe7Cp0dW3Hjjjdx+++28/PLLJCYmkp2d7fAazug9IM3T9uECf9Xpi7iH+hwdGcb95yzn65tPZPWsaPRGC//ak8u2VxPIr+vwSk5SOO7aKtoUEdTaXPTbrE/mR3DwSXHQM/VIWLFtYPveWZ8f7oWXjv5t0KOC8fMB0EcthNXiekyaDlkfIQjioAegVg1fXge/HRLaD30fs++zmFrRpjQEtTa97NPX/aYrSJ7xGeiQMJVKhSAIqFQqrFb3F4YGGlyNGPV6PVFRUYPynbUPp/9cXF2HYIrrzKfVJvBhUiVP/1hEp9FCaIiKa0+cz9+3LGJURJjHOUnhuGuraFNEUGuzPgve2CwuZu7FZbth8emD2Gf/Zv9bnzttFZzzPMw4CqoS6YycwhiVEV47AUaOh7s1v3N1tfD8ckd/295CP/9MxzrMPfD4THEhNcBxf4GtTwKKNqUiqLXpZZ++6Dd9OuNTUVHR71VeXm7/OpxRUlIiq324wF91+iKuM5+hISq2b5jHT7dv4qyV07DaBN6IKee052L4uaDB45ykcNy1VbQpIqi1OX01XPaZuLMzoJ+w6vdZoAHtV8Ht+XDlV3DFF3D9AZh1jLgWaN4JFNfrYYI4+0N3G7RXgyZe3CuoQdx/jcnL4OT7xO+/vpmGuF2OMRpyfx/0gDjj01EHNWmKNiUiqLXpZZ++7jddQfLAp7KykpkzZzJ37lyH18yZM6msrPRaYkOJnTt3snz5ctatWwdAa2sr8fHxWCwW+/RaTEwMY8aMIT09ndraWkpLSykoKKCpqYnk5GSMRiN1dXV22+7ublJSUmhoaKCoqIjS0lIMBgPp6el0dXU5+DWbzSQkJNDc3ExeXh7l5eVUV1eTmZmJTqdzsLXZbMTGxtLe3k5WVhZVVVVoNBpyc3NpaWkZMO/Ozk7S0tIc8m5sbCQ5OZmenh4H256eHnvehYWFlJSUUFdXR1paGp2dncTExDB16lSHvFtaWsjNzaWiooKqqiqysrLseQu/TZ/3ftXpdPa8Kyoq7HknJCRgNpsHzLuuro6SkhK6u7tpaGggJSVlwLyTk5NpbGykoKCA0tJSamtrHfLutbVYLMTHx9PS0kJnZycajcaed3t7O7GxsdhsNjunOCuFx89ZyP0bJzItKpza9m6ufV/NNW8nUFSl7Zd3V1cXarWa+vp6iouLKSoqsufd3d3toBOj0UhycjJNTU0Oeaenp6PX6x1se/NubW0lJycHjUZDZWUl2dnZjBo1iri4OIe8Y2Ji6OjoICMjA5VKRVlZGXl5eTQ3N5OYmIjJZLLbarVaDAYDarUarVZrz1ur1ZKamorBYHDwazQaSUpKoqmpifz8fMrKyqipqSEjIwO9Xu9ga7VaiYuLo62tjezsbCorK+15t7W1ERcXh9VqdeDo9XoyMjKoqamhrKyM/Px8mpqaSEpKwmg0OtgaDAZSU1PRarX09PRQXFyMVqtFrVb3y9tkMpGYmGi/1/rm3dHR0e9ei4uLo7293Z63RqMhJyenXx/Re0/o9XqnfURf/4f2EcXFxdSPOZKMDa9iPP1pEmbfCCqV8z6iTItu0tHE1EVASKhDHzFq1CiyCkqxjJwMQPd7F8J7Z6F9+zKsdVli5zftSGKFo7EccSpYejgidQfVSXvsfUR5kvi4vE0VCoChJhfbrgvgrVOYk/8qhQUFA/YRffu2hoYGp33ElClTBu0jNBoNbW1ttLa2utVHFBYWer2PyM3NddlH9OadmZlJdXU15eXl9nutN+++97KrPuLQe81ZHzFQ3n37CEEQyM7Opr293Wkf0XuvDdZH9N5rarWasLAwWX3EmDFjJPcRkydPdtpHFBYW4i4kf9QVGhpKfX09U6Y4Purb0tLClClThvVHXaWlpSxcuHBQvrP24TRl6+o6BFNcqT4NJgsv/FzCW7EVWG0Co8JD+MeZS7lq/TxC3Vz8LCWmu7aKNkUcztoclP/OVqg65ImYEdHQo4MtD8KJfwerGT67StzscOwsuDFBtPn+Hkh+FZZfAPl7+gdx9lGcm3Uo2gy+uF7Tphc5Pv2oq3ctz6FoaWlh9OjRUt0FFcxm5+c6uWofLvBXnb6IK9XnqIgw7t26jG9vOZG1c8ZhMNt48Jt8LnwlntxanddjumuraFPE4azNQfkTjujf2PObVqcdKX4NDYcLX8c0eiZ01Pz+OHxjvvh14RYYNam/n/T34cCTsPsa6G53noerPIc5FG3K43uzDrd3ENq2TXyyQKVScfXVVxMZGWlvs1qtZGdns2HDBq8lFoiYNGmAG19C+3CBv+r0RVxPfS6bPpYv/rKBN37JZ2dcDdk1Os57OY6rN8zn9tMXMyZy8FtLSkx3bRVtilC0OQC/d50PiOuIFp0BMU+JP09d+XvbiLGYVl5KRNIzkP8/OPrq3wc+U1fAmksh4SXx54VboPQnKPxWfAFMXgqb75Zch6LN4IvrNW0OYcy+cHvGJzo6mujoaARBICoqyv5zdHQ006ZN4/rrr+fDDz/0WmKBCFeLt4f74u5e+KtOX8SV4zMkRMXaqE5+vmMT566egU2Ad+IrOO25g/yYp/VKTHdtFW2KULQ5AL/vjM/s4+Dkf8IZj8FZz0CU4w78pRErxG/KfhGPy+hqAlQweQkctf13w/U3YxgzzzGg+h3xIzOJdSjaDL64XtPmEMbsC8lrfB588EH+8Y9/MGrUKK8lEShw9Rmh0Wh0mOmS0j6cPqt2dR2CKa5cn335B4ub+NeeXKpaDQBsWTaVB89fwcxxIz2O6a6tok0RijYH4NdlwhubxDcH2wyxL+edU35/6gtg/Dy49bfF0D8/DO2VcMGrGJs1RDbnibtKv7wOuhrhj+/Ckdv6+1S0qWhTJt8Vx6drfK666ipqa2v7vV9SUoJGo5HqLqiQnJwsq324wF91+iKuXJ99+ZsWT2bfbRu56eQFhIWo+KmggdOeO8hbseVYrLYBOd7KT9GmCEWbA/AnHAGq37r62ce55my6CyYuEl8g7v7ci1P/BX94C0LDSS6qhxUXwMhxcMw1YnvKm67r+OlBeO8csBgHbh+mULQpj+/NOiTP+GzatIk///nPbN++3eH9Dz/8kLfeeosDBw54LbmhhnJkhQJvobhBz31f5ZCqEc9RWj59LI9tW8ma2eP8m9gAULR5GCDjI7Cafh+guAtTF4SP+v2MsMHQUQ//OVLc8+eGGHEt0UCwWeGh347G+NMnsPSswX3mfYUt4WV+GncZJ19wlaJNBU7h0xmfjIwMTjjhhH7vH3/88WRmZkp1F1RQtl4XoWy97pq/eGoU/71+PU/+YSXRI8PJr+/gwlfiueOzLPbsO+D1/BRtilC0OQh/7eVuDXr6xYwYPeigx8F27HRY/tv5jYk7B7drrfj9e0u3Y7uuFoz639t3X01IrZrltf91mXcwQNGmPL4365A88FGpVOj1+n7v63S6oN7Dxx30bnDoaftwgb/q9EVcuT6d8UNCVFyybg4/37GJbWtnIgjwRXoN98R089z+YrqMlkG5UvNTtClC0aY8vhROP9v1N4tfc7+AhryB7XqfEgPQa39vr8+CF9fCJ5eKbb+dUA8wwtzudk6BDEWb8vjerEPywGfjxo08/vjjDoMcq9XK448/zoknnui1xAIROTk5stqHC/xVpy/iyvXpDn/SmEieu2QNX924gWPmjqfHYuPFn0s4+ZkDfJZajdU2+KfN7uanaFOEok15fCmcfrYzjxIflbdZ4NUN8Pm1YLP9bleXCZ9d+bt9R93vfn74J1iNoImFzkaojLebjTS39Q/eVil+bNYXFpN4+GqAQtGmPL4363B7H59ePPnkk2zcuJElS5Zw0kknARAbG0tHRwe//PKL1xILRMydO1dW+3CBv+r0RVy5PqXw184Zz+6/rOeTuEJeS9RS1Wrgri+yeSe+gvvPXs6Ji/rvU+Guf0WbIhRtyuNL4Qxoe+Fr8N7Z4sxO7uew7v+YO3uO2PbjfY62iS9D1HTmzTgV9v4+0KH8AOhq7D+ONjVi7umAtN0wepI4MPrxXoiIgj++LZ5S390GLx0Dc46HP30koeKhg6JNeXxv1iF5xmf58uVkZ2dz8cUX09jYiF6v56qrrqKwsJAjjzzSa4kFItrb22W1Dxf4q05fxJXrUypfpVJx9JQQ9t++kfvOWkbUiDAKtXqueDuZP7+XSmmj48fI7vpXtClC0aY8vhTOgLajJoiLm3v3Dnr3TKK+vFw83LQyrr/9vvtQZXwI9JmpyfwIqhIdzMLe3wo/3A1fXCsOegBMevGjsc5GKD8IhmZxM0WTwe0ahhKKNuXxvVmH5BkfgBkzZvDYY495LYlgwUBHdUhpHy7wV52+iCvXpyd8lUpFZFgo1208gj8ePYsXfi7hw6RKfils5GBxE5cdO4e/b1nExDGRbvtXtClC0aY8vhTOoLah4bD2Svj5QQBGNWWImyGCuOfPuv+D/15hN5+c9bL4zZz14oCn/ED/WM1FA8cSrFD8I4T32SurMV88pT7AoGhTHt+bdUie8QHxo60rrriCDRs22Pf0+eCDD4iLG2BEP4wQFRUlq324wF91+iKuXJ+e8Ptyxo+O4IHzVrDvto2ctnwqVpvAB0mVbH76AK8dLCNipHvn3ynaFKFoUx5fCsep7dxDji9K/W1/n6OughlrB+Zs/AdMWODwlsAgf+wmLYbN/xS/z/vqtx2mf0N95uB5+RGKNuXxvVmH5IHPF198wRlnnMHIkSNJT0/HaBQ3odLpdMN+FmigjRultA8X+KtOX8SV69MT/kCcIyaP4c2rjuGT647nyJlj0RstPPF9IRe9m83XWXW42m5L0aYIRZvy+FI4Tm1nHAWTlji+NyJa3Axx7ExYc4U4CJq0WGybvAyOOBnOf9mBIswcZOZm7IzfdqBWQdnPjh+NNZe6V0BbJfzyKHQ2ubb1AhRtyuN7sw7JA59HHnmE1157jTfffNNhQ6kTTjiB9PR0ryU2lNi5cyfLly+3Py7X2tpKfHw8FovFvndATEwMs2fPJj09ndraWkpLSykoKKCpqYnk5GSMRqP9M8iYmBi6u7tJSUmhoaGBoqIiSktLMRgMpKen09XV5eDXbDaTkJBAc3MzeXl5lJeXU11dTWZmJjqdzsHWZrMRGxtLe3s7WVlZVFVVodFoyM3NpaWlZcC8Ozs7SUtLc8i7sbGR5ORkenp6HGx7enrseRcWFlJSUkJdXR1paWl0dnYSExPDsmXLHPJuaWkhNzeXiooKqqqqyMrKsuctCILDV51OZ8+7oqLCnndCQgJms3nAvOvq6igpKSEsLIyGhgZSUlIGzDs5OZnGxkYKCgooLS2ltrbWIe9eW4vFQnx8PC0tLahUKjQajT3v9vZ2YmNjsdlsDhydTkdmZibV1dWUl5eTl5dHc3MznZ2d/fLu6upCrVZTX19PcXExRUVF9ry7u7sddGI0GklOTqapqYmCggImC228euF8bjk2milRETQZbPztkwy2PPkDKeXNxMfH09raSk5ODhqNhsrKSrKzs5kxYwZxcXH98u7o6CAjI4Po6GjKysrseScmJmIymey2Wq0Wg8GAWq1Gq9Xa89ZqtaSmpmIwGBz8Go1GkpKSaGpqIj8/n7KyMmpqasjIyECv1zvYWq1W4uLiaGtrIzs7m8rKSnvebW1txMXFYbVaHTh6vZ6MjAxqamooKysjPz+fpqYmkpKSMBqNDrYGg4HU1FS0Wi0REREUFxej1WpRq9X98jaZTCQmJtrvtb55d3R09LvX4uLiaG9vt+et0WjIycnp10f03hN6vd5pH9HX/6F9RHFxMfX19ajVarq6uhx04kkfMWPGDMl9xLhx49zuIyZNmjRoH0FYBDErn8Z6wh32frZ51hlU1NRTVV1N1vzr0W16hLQjbsZ85CUkL7wDQaUiptKCcdXvT351hY6zf593dJ9/rKNmEJNXi3Wm2GcLBd/Zm3oay9zqI4zvXQAxT9H20Z/pbKyiYtdNoKvt10fk5ubK6iN6+7a+v09XfcSh91rfPqK3b0tPT+93r/Xm3bePiIqKIjs7m/b2dqd9RO+9Nlgf0XuvqdVqJk6cKKuPmD17tuQ+YvHixU77iMLCQtyGIBEjR44UKioqBEEQhDFjxghlZWWCIAhCWVmZEBkZKdVdQEGn0wmAoNPpBmw/ePCgU76zdpPJJOzZs0cwmUyycgwEuLoOwRRXrk9P+O5yDEaLcNtb+4Rl//pemHv3t8Lcu78VbvwoTahq6ZLsU9Fm8MUNZG26bZv7pSDsGCu+dHXu+Yl9zs4x530rCDvGCrZnlghCT8fvvr7+m2j76RW/v9f7en2T85yMnYJQn92ft2OsILy5xXVNHkLRpjy+K46rv999IfnIiiOOOII33niDLVu2EBUVRVZWFkcccQS7du3iiSeeID8/37WTAIVyZIWCQERjRw/P7ivms7RqBAEiQkO4+oR53HTyQqJHyteSok0FPoPFBN/dDnNPgDWXusdRvwPf3gaA+e5aEr56kw1bLyZ8/Ex4IFq0WXM5XPAK/O8myPjQkT96CvyjZHD/750j7hc0GB7QuZengoCCT4+suO6667j11ltJTk5GpVJRV1fHRx99xJ133slf//pXj5MOBijHAohQtl6Xx5fCiYmJYcrYETz5x1Xs/dtJnLhwEiarjTdiytn89K/s/LWUlIpWfvz5oNfzDEYo2pTHl6pNlwiLIGb8xU4HPf38jJ3Zhx9J++gjYMwU8efN90JkNGz4m/hz5AB/4LoaQd8A8S9AS5ljmyA4H/T4EIo25fG9WYfkGR9BEHjsscd4/PHHMRjE/RIiIyO58847efjhh72WmD/gasRoNpud/kfsrH04/Vft6joEU1y5Pj3hS+EcaisIAgeKmnh0bwGljZ0OtvMnjWbFjLGsmBHNkTPFrxNGR7iMqWgzMOMGmzY9tevXbrPBt3+HaSsxr726vzZtNgj57X/2Xx+Dg0+K34+fJ25k2NNnxmbkBPh7jnhA66eXQZsG9PXOE76/CcIiXNYlFYo25fFdcXw646NSqbjvvvtobW0lNzfXvngp2Ac97iA1NVVW+3CBv+r0RVy5Pj3hS+EcaqtSqTh56RR+uPUkHt+2ki3LpjI9egQAFc1dfJtdz5M/FHLl2ykc9fB+Njz+M9ftUnPvhzH8lN+AVtfj8gmxYIaiTXl8Odr01K5fe0gInPciHHvdwISQPn+2Ivs84jxnPZx0p6Ntdyu0lEDSq+KTX64GPQANua5tPICiTXl8b9bh0QaGABEREURFRREVFcWYMWO8llAgY/HixbLahwv8Vacv4sr16QlfCmcw27DQEC49dg6XHiseB1BSVU99Txi5dTry6jrIq9WhaTFQp+uhTtcDwOdFagAmjYlg+YxojpwxliXTopgwMow6AzR3Gpk8NpSwUI+29woIKNqUx/eGNqXayaqz78Bn7ExYewXs/5ejTV0GFH/vvs+qRPFU+rjn4eT7YNxs93iNBTBuLkSMGrBZ0aY8vjfrkDzwsVgsPPjgg7z44ot0dopT7WPGjOGWW25hx44dQT9V7gwNDQ1MmtT/PCV324cL/FWnL+LK9ekJXwrHXVuTvpWNK1awcfFk+3sdPWYK6jrIresgIb+Smq4QSps6ae40EVPcRExx3/1Lwngy6yAqFUSPDGfi6Agmjo5kwugIJoyJYOLoCCaMjmDimEhmjhvJUXPGBeRu0Io25fF9oU2f9pt9Bz5jpopHZhyK3xZKD4jzX4H/3ej4XmUCZH0C2hxxEHRr1uB8ixHCIqH0Z/hwGyw5Cy79ZEBTRZvy+N6sQ/LA55ZbbuHLL7/kqaeeYv369QAkJibywAMP0NLSwquvvuqVxAIRI0eOlNU+XOCvOn0RV65PT/hSOO7aDmQ3dkQ4xx0xkeOOmMjJMwSOOOIIesxWCrV6cmvFmaGypk5aOo1o2zoxWFUIArQbzLQbzJQ1dQ0a75oT5rHj3BVu1zFUULQpjz9U2vQ0Zj/0XdwcNVX8etpDkPw6RE2HWnUfYxUOZ4IBTFkKV38H/7sZVl0srheqTBA/IgNxTZDVIq4bGjFWPI5DEKBoL7SUwk8PwrJz7SfNU7TXN3XKwOGoTVeQPPD5+OOP+fTTT9m6dav9vVWrVjF79mwuvfTSYT3wcTWbNZxnu/rCX3X6Iq5cn57wpXDctXVXmyPCQ1kzexxrZo+zt/Uubj7jzK10mgVau0y0dJpo6TLav2/tEl9NnUZSKlp5N17DKUuncNKiyYNE9A8Ubcrj+1ObHqHvwGfMNPHrCbeKrx/vcxz4hIaLR2I0Ffz+3qTF4qzRrZnio/exz/4+6OnFwxPFr8svgIvfh7wv4fM//96ev8d1ntWpjG6thyOOkFCcd3A4atMVJH+YHxkZybx58/q9P3/+fCIivL8SPpDQ0tIiq324wF91+iKuXJ+e8KVw3LX1hjZDQ1RMGhPJ4qlRrF8wkXNWzeCq9fO47bTFPHzBkey8/Cg+u2E929fPBeCuz7PRdZvdym+ooGhTHj9QtTkoIvusL+195L0X4YestVGFiLM7297qw+/zUVlYBETNGDxW/h5oLoHCwWd1AHHGqC/MPfD2FqZ+e6XjE2dDhMNRm64geeBz88038/DDD9vP6AIwGo08+uij3HzzzV5LLBAxf/58We3DBf6q0xdx5fr0hC+F467tUGrz7q1LmTdxFPW6Hh76JrA2LFW0KY8fdNoM6TMLMGaqY1vfQQ2IewCNngirLoJrfxIfcz8UY/sMfMJH/X6WWC92XwO5nzvP6d2t4tqf5lJIeRP6niw/2Dli2hzx4zmLyblvD3A4atMVJA98MjIy+Pbbb5k1axZbtmxhy5YtzJo1i2+++YasrCy2bdtmf3kTMTExnHvuucyYMQOVSsWePXsc2js7O7n55puZNWsWI0eOZPny5bz22mtezSEry8kiNzfahwv8Vacv4sr16QlfCsdd26HU5qiIMJ69eDUhKvgivYZ9eVqv+ZYLRZvy+EGnzYkLxcNNV13S/2mqddfCUdvhss/gqq9hfZ9/zGevg3Fz+vuL7rN54qa7xfVCfdEwwGBpIDw+C14+GvbeCQee+P39voOgXpi74bUT4fu74JHJkPGRezHcxOGoTVeQvIHhNddc47btu+++KzmhwfD9998THx/P0UcfzbZt2/jqq6+44IIL7O3XX389v/zyC2+99Rbz5s1j37593HjjjXz55Zecd955bsVQjqxQcDjCU20+/n0Brx8sZ9KYCPbdtsm+WaICBd7CkPeb+/4FCS+K31/+BcxdL258WH7AO/5PfwQ23OL4XksZvHSU43vKsRmS4dMNDN999123X97E1q1beeSRR7jwwgsHbE9ISGD79u1s3ryZefPmcf3117N69WpSUlK8loNyZIUIZet1eXyvHwvghp0vrt3tpy1m8dQxNHeauH9PTkBsjKhoUx5/uGjTYyw6/ffvJy8W9/O56n/9Byuewqjv/153m+PPYb89vWTq8sqaoMNRm64gecbnUBw8eJCuri7Wr1/P+PHjvZWXU6hUqgFnfDIyMtizZw8zZszgwIEDnHfeeXz33Xds3LhxQD9Go9FhrVJHRwezZ8+mubl5wBGjzWYjJGTwsaKzdrPZzP79+znttNOCfsbH1XUIprhyfXrCl8Jx19Zf2syr6+CPrydjsQk8f9FKzlk1XRLf21C0KY8/nLTpEQSBkB/vBlTYzngCfturKiTpZUJ/fkC2e+uxN2A77dE+b5gJ/eJqQkp+/D2FyLFYbi8m7JVjwdKD5YYEGDnO45iHizY7OjqYNGmSWzM+bj/O/uSTT9LZ2Wk/mkIQBLZu3cq+ffsAmDJlCj///DMrVvhnb4+XXnqJ66+/nlmzZhEWFkZISAhvvvnmoIMegMcff5wHH3yw3/v79u1j1KiBd9+Ui/379/vErwIFcuGpNk+boeL7mlDu+zIbfXkG0conXgq8jKHtNzeLX77/fbfnaEMIx4eNZYSlo5/1vhXPsajhW6INlUwwlPVr74ua0gKq255jXFcFFZNP5Yim/ayo+9HBxmI28/M3n3GmrgqAzC+fo2788fJKAkJsZlbWfEBT1HLJ/kJsJkJsZixho2Xn4Sv0nh3qDtye8TnqqKO4++67ueSSSwDYvXs327dvZ//+/SxbtoyrrrqKUaNG8dlnn3mWtQQMNOPzzDPP8Oabb/LMM88wd+5cYmJiuPfee/nqq6/YsmXLgH6kzvjodDqio6MHzctZ+3Ca8XF1HYIprlyfnvClcNy19ac2zVYbF7+RQm5dB5sWT+LNK9b6bVdnRZvy+MNNm95G+KP9dw4239csfmOzEpL2NqH7/jkoXxgzFVVng9MYQmQUlqv3Ef66uEGw9cynsR3t/traQ9F7fVWF3xL2xdVizv9sss9muYOw55egMrRgvrMCIqMCUps+mfGpqKhg1apV9p/37t3LH//4R0444QQA7r//fi666CJ33XkV3d3d/POf/+Srr77i7LPPBsRNFTMzM3nmmWcGHfhERkYSGRnZ7/3w8PABb7La2lqnW2a7anfmO5jgTp3BEleuT0/4Ujju2vpTm+Hh8NwlazjnpTgOFjfzVZaWS9YN8MTMEEDRpjz+cNOmzzFmap+8wmHDTWDtgZ8PeRosMhqMOpeDHgCVUU94S6H951BLF6Hh4fDLIxDztLgf0WWfwaLToLUCOmrFR+/LfhE3bgx1vE726xv2+5/7cJMOxri5+ajNBgZxD53w1hKYc9z/t3fm4U1V6R//pntpS7ELhbIUKGtllU12UFRQEMYRdxFE/TnjOriMMuPgjDNug+PKoA4OqOOojKMw4gaydU+b7nubtumeNk2aNG3arPf3x6WhoWmTm5vbtOn7eZ4+tDnv95z3hG9OT8+995wh6U0u/nD6IpvJZLKZJKSlpWHlypXWn2NjY9Ha2up0w+7EaDTCaDT2uf7n6+sLi8XitnYc3cM0WPc4eRpP9VOIdvnW6Yqei8bZWE97c2ZMGJ6+nt3z5E/fFqNO5fyyszshb/LTe6M33UrQGPbf2KuANU8Bu+0cfhpiZ0IRyXHH5q96rfCcfxWwmNlJDwAwFuCzW9nv31kIHL0J+McG4OxLQP6Xfaqyvr+9L+50q53Pxdjr6Bq/ANs6XURobzrC6YlPfHy89a7q2tpalJeX29w/U19fj8jISLcldjkdHR3Izc1Fbm4uAHYFKjc3F7W1tRg9ejTWrVuHZ555BufPn0d1dTWOHj2KTz75pN+nwFzB0STKnZOsoYyn+ilEu3zrdEXPReNs7FDw5p7V07B0yhXoNJjx7Ff5sFgG/ykv8iY/vbd6023s/gFY+iBw5xfAtX8AIuP7xvgFXfp+Mnu5Cuufd71NUzfw8oS+r3/zcN/X1HV9XrK+v6ZLt3XgvSXs/kGOKP0OKPz60s8+frZ1uojQ3nSE05e6HnnkETz66KNISkpCeno6VqxYgYSEBGv52bNnsWjRIrcldjkSiQQbNmyw/rx3714AwH333YejR4/iiy++wPPPP4+7774bKpUKcXFx+Mtf/oKHH7ZjDhfpOY3e1XJvwVP9FKJdvnW6oueicTZ2KHjT10eEAzsWYNNbSUirUuLjNBl2rxrcnZTJm/z03upNtxGTANx0YOCY3ifE3/M1+0j6KJ6LAiY7k5Q8O6fAX3iN3STx2j+wbQaFX3p/Td22scUngAV32L6mlbMrVj6+7GP2X9xlW25mj6gZ6t50hNMTnwcffBC+vr749ttvsXbtWuzfv9+mvLGxEffff38/av6sX79+wH1Cxo0b5/a9gy5n/PiBH9V1VO4teKqfQrTLt05X9Fw0zsYOFW/GRYZg301z8MLxQrz6QynWzoxGfHSoY6GbIG/y03uzNweNadewO0nHXMnuJn1xR2nd8t9glLIAuHY/cOaPgPRnARpngKJv2C8AeOgCxo+fxH5/+cSH6bWCYjYC0jPA57cDV/4C2HHU/h5CFyc+Q92bjuD0IP3999+Pb775BocOHcK4ceNsyv7+97+79bLSUKS0tJRXubfgqX4K0S7fOl3Rc9E4GzuUvHnP8slYMyMKepMFT/8nDybz4F3KIG/y03u7NwcFHx/glg/ZG417kTP6OuCe/wLj5wN3fG57L9CCO4HHc4Dp17k3lw/XoTIvDehQAPWZtmU9GycqK4HXprCTHuDSpMls6lufmT1LbKh70xG8NzD0JhxteW0ymeDn1/8i2UDl3nRkhaP3YTi1y7dOV/RcNM7GDjVvNqq7cMObidDqTZgaFYJAPx8wDMCAufgvuxcYAwAMEBLohz2rp2Lbwlhej8KTN/npR4I3PUWffn62A6hg98HDE3nAFVOAk3sByUfOVzptA1B1bsAQ801vwfe7J+0X7lcDfxxj/3V5PvCBnX3wJl0NyxVT4LPiEXZ36YwPgRteBkZfXJEx6QG/vk9L90YIbwp6ZMVIJjU1lVe5t+CpfgrRLt86XdFz0TgbO9S8GTsmGH/cxm5mWt3aiVK5FmXNWpQ3d6CipQPSlg5UKjpRpehEVWsnCho0ePLLXNx9WIxKhfPX8s0WBsckdbj7cDr+cKIQR08meuSmavKm63EjdtycveXS9/4XNwY097oB+YVW9pJTb56+7HR3J25Q7nfSAwDZn9h//dzL/R+WWpcOn/wvgC/vBo5sAoq+Bk7+hn1q7N93AH8eC/zwW1uNoZNdWbqI0N50BK349IIOKSVGIkJ6s6BeA3WXASKIIBIBIgAQweZnkUiEjGol3j0rhd5kQYCvD/5v3TQ8smE6gvx9+607sVyBl78vQanc9vyjCWOCsXVBLG5eEIs548M8tpkiwR+vHjelZ4B/3cJ+v6+RPResKQ/4cD37NNju74GGbPZRdYB9Wux3ctsVmhnXX1o1mr0FKD05mD24RPhkYOb1QObhS6/1Pmj1xYsbDy68B9h+UJAUaMVHIIbVYXsCQgdB8tOPpIMg500Mx5oZ0Vg9Iwqrpkdh5fQorIyPwor4SFw9LRLLp0Vi2dQIPHrNDJz+zTpsmBUNg9mCd89Kcf2biThX1tKnzpKmdtz7kRg7/5mBUrkWo4P88OTGGbhl0QQE+QIN6i68f6ESN76ThOveTMS7ZyrQ2qG3k537IG+6Hjdix83oWZe+7zmYdPwC4LFs4I6Lqy0xcy/FmLr77rZ87R+A6NnATW9c0ngCTa3tpKc3qqpL3+f+C8AwO6TUaDQiODgYubm5mDt3rmPBMMPRjLGjowOhof0/oTJQuTf95eLofRhO7fKt0xU9F42zsd7iTYZh8FORHH/8thhNGvYplM1zx+EPWxMggghvnCrDV9n1YBjA31eEnSum4LFrpmPMKHZjtdY2DTLqdfhfbiPOlrXAYGJvrB4V4Ivdq6bgoTXxCB/l/j6SN12PGy7e5IvdfpacBPyDgenX9i/81y/ZJ8AmLgUe+PnS6gnAXg7rvVPzsfuA4uPs6tDlT3ENNlvfBhbvss0XAF7UCOJNwVZ8/P39MXnyZJjNZi4yr6GsrIxXubfgqX4K0S7fOl3Rc9E4G+st3hSJRNg0dzxO712HB1ZPha+PCD8UyrHxjQtYf+Ac/pPFTnpumj8eP+9dhxe2JFgnPQBQUyXFjfPG4/17F0Py+434663zMX9iOHQGMw6eq8Sa18/ivbMV6NTbeWKFB+RN1+OGizf5Yrefc7YMPOkBgLu/Am7/DLjj3+zPCdsvlV12PAV+8QHwf0nAs1XwON8+YbtbdC+E9qYjOF/q+t3vfod9+/ZBpVK5LYnhwuWP8HMt9xY81U8h2uVbpyt6LhpnY73Nm6GBfvj9lgScfGw1rpo8Bp0GM7qNFiyJuwJf/3olDt51FeIi+54U3bufo4P8sWPJJJx4ZBU+uHcxZsWEob3bhAOnyrH29XM4nFSFbqN7/ogjb7oeN9y86Sou91MkYidIoWPZn9c+w67oXPa4PADAP4h9XD4gxPYeG09x+RNjweyxE0J70xGcn5V87733IJVKERsbi7i4OISE2A4+2dnZbktuqNHVNfAd9I7KvQVP9VOIdvnW6Yqei8bZWG/15pzxo/HVwyvxQ6EcwQE+2DBr7IA3K9vrp0gkwg1XjsPGOTE4md+IN0+XQ6bU4c/fleD7giZ88dAKBPjxu92RvOl63HD1Jlfc1s9xc4Hnah0+Mg4AuPIW9qmrgVi8G4iaAfzU/6nybqOrDfjnJuiXvwxMsHMMx0BSN/qE86d9+/btePrpp/H888/jrrvuwrZt22y+hiMHDx5EQkICli5dCgBQqVRISUmByWSy3lCVmJiI9vZ2ZGdno6GhAVKpFCUlJVAoFBCLxdDr9cjLy7PGdnV1ISMjA83NzSgrK4NUKoVOp0N2djY6Oztt6jUajUhNTUVrayuKiopQVVWFuro65ObmQqPR2MRaLBYkJSVBrVYjLy8PtbW1kMlkKCwshFKptJt3R0cHsrKybPJuaWmBWCxGd3e3TWx3d7c179LSUlRUVKCxsRFZWVno6Oiw5ts7b6VSicLCQlRXV6O2thZ5eXnWvBmGsflXo9FY866urrbmnZqaaq338rwbGxtRUVEBmUyG5uZmZGRk2M1bLBajpaUFJSUlkEqlaGhosMm7J9ZkMiElJQVKpRKVlZWQyWTWvNVqNZKSkmCxWGw0Go0Gubm5qKurQ1VVFYqKitDa2orCwsI+eXd2dkIikaCpqQnl5eUoKyuz5t3V1WXjE71eD7FYDIVCYZN3dnY2tFqtTWxP3iqVCgUFBZDJZKipqUF+fj7UajWSk5P75N3e3o6cnBzI5XJUVlZa805LS4PBYLDGyuVy6HQ6SCQSyOVya95yuRyZmZnQ6XQ29er1eqSnp0OhUKC4uBiVlZWor69HTk4OtFqtTazZbEZycjLa2tqQn5+Pmpoaa95tbW1ITk6G2Wy20Wi1WuTk5KC+vh7V1VWY6teGeZE+1s9a71idTofMzEzI5XLU1NSgvLwccrkcEonEJu+U5CRsTojGSysD8fvrpyLEX4TsWjX+9HUWcnJy0N7e3uezlpycDLVabc1bJpOhoKCgzxjR4wGtVjvgGNG7/svHiPLycjQ1NUEikaCzs9Pm/36wxojm5manxwiFQtHvGNE77552+xsjenxob4yQyWRoa2uDSqVyaowoLS11+xhRWFjIa4zoybv3/6ejMeLyz1qfMaK51TpG2Mu7Z4zQ6i+taJrHsMfHMFc/iowV/wAi4lE97R60r/sTSgyDuOpWm4bIC89xHiMu//z0HiMqKys5bXBIj7P3wtHNUS0tLRg7dmy/+oHKvekmPUfvw3Bql2+drui5aJyNJW+ycHlvfyhowq8+y4ZIBHy2ZzlWTo8alHYHq07y5tDCI+Pmd09detrqkUwgNNp6ucmGNhnw9oJBS4vxC4bo93JOGkfvn+CPs6vVahw+fBjPP/+89V6fnr9yvJnq6mpe5d6Cp/opRLt863RFz0XjbCx5k4VLPzfPG487l00CwwC/OZaLtk7DoLQ7WHWSN4cWHumnsdeTXWHj7E96gEsbKALAonvZIzQuJ26V29IS2Tt01QHufP84T3zy8/Mxc+ZMvPbaazhw4ADUajUA4Ouvv8bzzz/vtsSGIgsWDDwjdlTuLXiqn0K0y7dOV/RcNM7GkjdZuPbzhS0JmBYdguZ2PX773/wBD0J2Z7uDUSd5c2jhkX7GrWD/HTcPCBpgFeTiQaps7HzgF+8Dtx5hD1uduQlY/itg13f2tde8AMy4Adj9I7Dre/flfhnufP84T3z27t2LXbt2oaKiAkFBQdbXb7zxRq/fiCojI4NXubfgqX4K0S7fOl3Rc9E4G0veZOHaz1EBfnjnjkXw9xXhVHEz/p1ROyjtDkad5M2hhUf6ufBuFM/9LbDj44HjejZQBC6d2j73Fvaw1bu+BDa/2nfzxB7m3QrcfYydZMUudCotfUCEU3G9cef7x/ken/DwcGRnZyM+Ph5hYWHIy8vDtGnTUFNTg1mzZqG728ObJvGAjqwgRiLkTeBwUhX+/F0Jgvx98O2jqzEjJszTKREgbw4qPRsN3vAKsOLX9mNS3gFOv8A+Up/4V/a1Z6uBURcnMmYj8JKde+VWPgakvnvp5981s4/euxFB7/EJDAxEe3t7n9fLy8sRHR3NtbphBW29zkJHVvDT07EAwuFqP+9fNRVrZkSh22jB41/kct7fh7zpehx5c6i1O8BayKrHgaelSPRdA8RfC8zcbHvfkI+dHXLm3w5c/2fbnNK4r9547MgKAHjggQegVCpx7NgxREREID8/H76+vti+fTvWrl2Lt956y23JDTaOZozd3d02l/e4lHvTXy6O3ofh1C7fOl3Rc9E4G0veZOHz/9mi7cbmt5Kg7DQgIiQAV8aORkLsaCSMH40rY8MxNSoEvj72l/vJm67HkTeHSLs9Kz49R024Wmfvw1cBYO2zwDW/u1T/2mfRvfIpt3tT0BWfN954Ax0dHRg7diy6urqwbt06TJ8+HWFhYfjLX/7CtbphRX5+Pq9yb8FT/RSiXb51uqLnonE2lrzJwqefY8OC8ObtCxEa6AdVpwFJFa344EIVnvgiFxv/dgFz9/+EX/w9Bb8/XoDPM2qRV6eG3mTm3W5/kDe9iyE/bm56lT3tff4d/Oq8/AiO1U+y/247CEzfCKx6XHBvOoLzzs3h4eE4ffo0kpOTkZ+fj46ODlx11VXYuHGj25IaqsTFxfEq9xY81U8h2uVbpyt6LhpnY8mbLHz7uXZmNCS/34hSuRbFje0oatSguKkdpU1adBnNyKlVI6dWbY2PCg3EC1vmYPnkyU7VX9igQZOmGxvnDLwDNUDe9DaG/Lh59a/YL3fWecth9vgMAFh0D/vFRe9Km07AeeLTs9y0evVqrF692m2JDAfa2toQExPjcrm34Kl+CtEu3zpd0XPROBtL3mRxRz+D/H2xcNIYLJw0xvqa2cKgurUTxU0XJ0ON7Shs0KC1Q48nvsjFVbHBePPuMLvnhzEMg8SKVnyYWIkUqRIA8OTGGXhy40xB+0LeHFqMqHFz29+BmlTgyl+4LSd39oPzxGfMmDFYtmwZ1q1bhw0bNmDFihUIDg52LPQCfH19eZV7C57qpxDt8q3TFT0XjbOx5E0Wofrp6yPC9LGhmD42FDcviAUA6E1mfHihCu+ekyK7sQvXv5mIx6+dgaVTIqDqNEDVaYCyQ4/vC+UoaWq31mO2MHjr5wpMi75UlxB9IW8OLUbUuLnobvbLVb2bNP3BeeLz888/IzExEefPn8ebb74Jk8mEJUuWYN26dVi/fj2uu+46tyU31Lj8QFau5d6Cp/opRLt863RFz0XjbCx5k2Uw+xno54vHrp2BLQti8cyXEkjqOvDXn8rsxo4K8MWdyybj/tVT8XGqDB8mVuHp/+Rh0hXBWDTZ/m665E3vgsZNfnp39oPzzc2rV6/Gvn37cOrUKajVapw7dw7Tp0/H66+/jk2bNrktsaFIU1MTr3JvwVP9FKJdvnW6oueicTaWvMniiX5OjQrBc8tD8ObtCzB9bCimRoXgqsljsHFODG5bMhH7bpyNtOeuxQtbEjBhTDB+u2k2Ns6JgcFkwYOfZKFBzW7fb7EwUHbo0aDuQn2bDrnSOrRoXd8Xjbw5tKBxk5/enf3gvOIDsHv2nD9/3vql1+uxZcsWrF+/3m2JDUVmzZrFq9xb8FQ/hWiXb52u6LlonI0lb7J4qp+zZ8/GktBQ/GLRRIexvj4ivH3HQvzyUCpK5VpsfTcZ/r4itHYYYLZctrvI92cwf2I4diyZhJvnxyJ8lPOPdJM3hxY0bvLTu7MfnFd8JkyYgKuvvho//vgjrr76avzwww9obW3FN998gyeeeMJtiQ1FsrOzeZV7C57qpxDt8q3TFT0XjbOx5E2W4eLNkEA/fLRrKaJCA6HqNKC5XW+d9AT4+SDQzwcBPoCPCMiv1+CF44VY+vLPeOdMhWA5cdWQN7kxXLw5GHUK7U1HcJ74REdHQ6fTQS6XQy6Xo7m5GV1d3E9aHUocPHgQCQkJWLp0KQBApVIhJSUFJpPJultkYmIiFi1aZD2FXiqVoqSkBAqFAmKxGHq93lpfYmIiurq6kJGRgebmZpSVlUEqlUKn0yE7OxudnZ029RqNRqSmpqK1tRVFRUWoqqpCXV0dcnNzodFobGItFguSkpKgVquRl5eH2tpayGQyFBYWQqlU2s27o6MDWVlZNnm3tLRALBaju7vbJra7u9uad2lpKSoqKtDY2IisrCx0dHQgMTERa9eutclbqVSisLAQ1dXVqK2tRV5enjVvhmFs/tVoNNa8q6urrXmnpqbCaDTazbuxsREVFRUYO3YsmpubkZGRYTdvsViMlpYWlJSUQCqVoqGhwSbvnliTyYSUlBQolUpERERAJpNZ81ar1UhKSoLFYrHRaDQa5Obmoq6uDlVVVSgqKkJrayv8/Pz65N3Z2QmJRIKmpiaUl5ejrKzMmnfvz0piYiL0ej3EYjEUCoVN3tnZ2dBqtTaxPXmrVCoUFBRAJpOhpqYG+fn5mD9/PpKTk/vk3d7ejpycHEybNg2VlZXWvNPS0mAwGKyxcrkcOp0OEokEcrncmrdcLkdmZiZ0Op1NvXq9Hunp6VAoFCguLkZlZSXq6+uRk5MDrVZrE2s2m5GcnIy2tjbk5+ejpqbGmndbWxuSk5NhNpttNFqtFjk5Oaivr0dlZSWKi4uhUCiQnp4OvV5vE6vT6ZCZmQm5XI6YmBiUl5dDLpdDIpH0ydtgMCAtLc36Weudd3t7e5/PWnJyMtRqtTVvmUyGgoKCPmNEz2dCq9UOOEb0rj8iEPjzmhD8bft0vLd1Iv59+1QkPboI/94ejezn1+LwpjBk/G4j7poTiFkxoTCYLPjb6XK89b90p8aI+fPncx4j4uPjnR4jZs2a1e8Y0Xts8/PzG3CMWLNmTb9jhEwmQ1tbG1QqlVNjRGlpqdvHiMLCQl5jRE/evT/LjsaIyz9rA40R9vLuPUbExcUhPz8farV6wDGi57PW3xjR81mTSCSYOXMmrzFi0aJFnMeIVatWDThGlJaWwmkYF2hra2NOnDjB7N27l1m8eDETHBzMrFixgtm3b58r1Q0ZNBoNA4DRaDR2yy9cuDCgfqByg8HAHD9+nDEYDLxyHAo4eh+GU7t863RFz0XjbCx5k8VbvWmxWJg3TpUxcb89ycz43fdMbm2bIDmRN4XDW705WHpHGke/v3vD+ciK3iiVSpw/fx4nTpzA559/DovFArOZ2xk3QwlHW14bjcYBt00fqNybtl539D4Mp3b51umKnovG2VjyJos3e9NiYfDQp1n4uaQZ40YH4dvHViM6LNCtOZE3hcObvTkYekcaQY+s+Prrr/H4449j/vz5iImJwa9+9St0dHTgjTfe8PprtZmZmbzKvQVP9VOIdvnW6Yqei8bZWPImizd708dHhDdvX4D46BDI27tx/9FM5Nap3ZoTeVM4vNmbg6F3Zz84P9X18MMPY+3atXjooYewbt06zJs3z23JDHXo6QQWejqBn56enBEOb/dmWJA/Pty5BNsPpqCgQYPtB1Nwzeyx2LN6KqaPDUV0aCB8Lh6kSt4cWni7N4XWe/SprpaWFnz11Vd49NFHR9SkB6D9KHqg/Sj46WmvFOEYCd6Mjw7F94+vwa2LJ8LXR4SzpS24+7AYy18+g9l/+BH3H81Eh95E3hxijARvCqn3+D4+ZrMZx48fR0lJCQAgISEB27Zt8/qtx2kHUhbagZSfnnbHFY6R4s1JEaNwYMcCPLphOg6ekyKtSokmTTcMJgvOlrbgt1/l46kVYwAA0pYOvP5jKeLHhmLbwljMHtf//Q/kTeEYKd4USu/OfnCe+EilUtx4441oaGiwLj298sormDRpEr777jvEx8e7LbmhBp05wzKizpwRQE/nIQnHSPPmlKgQ/HXHAgCAyWxBepUKu49m4LuCJkwJY3Cdvxq7j2SgTWcEiptx6HwlFkwag3/sXIyxYUG8+kHe5MZI86a79e7sB+dLXY8//jji4+NRV1eH7OxsZGdno7a2FlOnTsXjjz/utsSGIm1tbbzKvQVP9VOIdvnW6Yqei8bZWPImy0j2pp+vD1bPiMLvb0oAABxKk+Ouf6SjTWfEvAnhuD4hBgG+PsirU/e7ESJ5UzhGsjfdoXdnPziv+Fy4cAHp6emIiIiwvhYZGYlXX30Vq1atcltiQ5EpU6bwKvcWPNVPIdrlW6crei4aZ2PJmyzkTWDnijjk1LbheG4jdAYzVk+Pwvv3LkZooB/SKpW48x/pOCapx2PXzEDMaNtVH/KmcJA3+end2Q/OKz6BgYE2O8r20NHRgYCAALckNVTJy8vjVe4teKqfQrTLt05X9Fw0zsaSN1nIm4BIJMLLt8zDilg/7F41BR/tWoLQQPZv3KunRWBJ3BUwmCz4R2IVrzbJm9wgb/LTu7MfnDcw3LlzJ7Kzs/HRRx9h2bJlAACxWIwHH3wQixcvxtGjR92W3GDjaAMkhmEgEon61Q9U7k0bcTl6H4ZTu3zrdEXPReNsLHmThbzpWH+urAW7j2Qi2N8XKc9dg2B/X2i6jBgXHkTeFBDyJj+9I42gGxi+8847iI+Px4oVKxAUFISgoCCsWrUK06dPx9tvv821umFFUlISr3JvwVP9FKJdvnW6oueicTaWvMlC3nSsXz8zGnMnjEaX0YxtB5Mx/48/YcWrZ/Cv9BrypoCQN/np3dkPl4+s6DnIDgDmzJmD6dOnuy0pT0ErPs5Bf7nw09Nf1cJB3nRO/2NhEx7+V9+d9l+5ZR7uXDbZrfkxDANFhx5jggMQ4Nf3b23y5vBrdyiOm4Ks+FgsFrz22mtYtWoVli5disOHD2Pjxo3YunWrV0x6nIH+cmGhv1z46emvauEgbzqnv+HKcXj91vl4+RfzcO7p9dizeioAYN/XBXjjVBnaOg1uy+/wiXNY8cpZPP91Aa96hjvkTX56d/bD6YnPX/7yF+zbtw+hoaGYMGEC3n77bTzyyCNuS2Q4sGDBAl7l3oKn+ilEu3zrdEXPReNsLHmThbzpnF4kEuG2JZNw1/LJmBoVgt/fNAe7Vk4BA+Dds1KsfPUsXvm+BN3G/g+ddja/DPUomC0Mvs1rhLbb6HI9wx3yJj+9O/vh9MTnk08+wd///nf89NNPOH78OL799lt89tlnsFgsbkvGUxw8eBAJCQlYunQpAEClUiElJQUmkwmJiYkAgMTERJSUlCA7OxsNDQ3WS30KhQJisRh6vR4//fSTNbarqwsZGRlobm5GWVkZpFIpdDodsrOz0dnZaVOv0WhEamoqWltbUVRUhKqqKtTV1SE3NxcajcYm1mKxICkpCWq1Gnl5eaitrYVMJkNhYSGUSqXdvDs6OpCVlWWTd0tLC8RiMbq7u21iu7u7rXmXlpaioqICjY2NyMrKQkdHBxITEyGTyWzyViqVKCwsRHV1NWpra5GXl2fNm2EYm381Go017+rqamveqampMBqNdvNubGxERUUFUlNT0dzcjIyMDLt5i8VitLS0oKSkBFKpFA0NDTZ598SaTCakpKRAqVQiOTkZMpnMmrdarUZSUhIsFouNRqPRIDc3F3V1daiqqkJRURFaW1tx+vTpPnl3dnZCIpGgqakJ5eXlKCsrs+bd1dVl4xO9Xg+xWAyFQmGTd3Z2NrRarU1sT94qlQoFBQWQyWSoqalBfn4+ioqKkJyc3Cfv9vZ25OTkIDs7G5WVlda809LSYDAYrLFyuRw6nQ4SiQRyudyat1wuR2ZmJnQ6nU29er0e6enpUCgUKC4uRmVlJerr65GTkwOtVmsTazabkZycjLa2NuTn56Ompsaad1tbG5KTk2E2m200Wq0WOTk5qK+vR2VlJYqLi6FQKJCeng69Xm8Tq9PpkJmZCblcjrS0NJSXl0Mul0MikfTJ22AwIC0tzfpZ6513e3t7n89acnIy1Gq1NW+ZTIaCgoI+Y0TPZ0Kr1Q44RvSu//Ixory8HE1NTZBIJOjs7LT5v3dljCguLnZ6jOjs7MSW2C48vCAQ0yMD0WU044PEKtxw4DSK65V2x4i8vLx+x4ieWLm6E2fLWgEABrMFXyYV9Rkjqqur+x0jZDIZ2traoFKpnBojSktL3T5GFBYW8hojevLu/f/paIy4/LM20BhhL+/eY4REIkF+fj7UavWAY0TPZ62/MaLnsyaRSJCbm8trjCgpKeE8RlRVVQ04RpSWlsJpGCcJCAhgamtrbV4LDAxk6urqnK1iyKPRaBgAjEajsVteU1MzoH6gcoPBwBw/fpwxGAy8chwKOHofhlO7fOt0Rc9F42wseZOFvMlPX1NTw1gsFuZUkZxZ9KdTTNxvTzIJL/zAVCk6bOK03Ubm4A85TKu2e8D6/pFYycT99qT16+FPJZzyJG8OzXaH4rjp6Pd3b5xe8TGZTAgKst3syt/fH0Zj36VLb8Vs7n/Z15lyb8FT/RSiXb51uqLnonE2lrzJQt7kpzebzRCJRLguIQY/PLEGCyaGo9NgxruX7fT8xqkyvH6+AVvfTUZJU7v1dYPJgjqVDgB7M+oxSR0A4I6lkwAA58sU0BlMvPMcjpA3+end2Q+nd25mGAa7du1CYGCg9bXu7m48/PDDNoeHff31125LbqjR2dnJq9xb8FQ/hWiXb52u6LlonI0lb7KQN/npe2tiRgfhpe1zcfN7KTie24DHr52BKVEh6DKY8VVWPQCgUdONXx5Kxcf3L8PSKRHY900Bvsqqx+GdSxAZGoDy5g4E+Irw/I1zkCxtRX1bFy6UKbB53njk1akxfWwoeXMYtutpb/LF6RWf++67D2PHjkV4eLj165577kFsbKzNa97M+PHjeZV7C57qpxDt8q3TFT0XjbOx5E0W8iY//eWa+RPHYMOsaFgY4L1zUgDAt/mN0HabMCE8ECvjI6EzmPHHb4vQoO7CNzkNAICD56XW1Z6NsyIRHuyPjXNiAABpVUqkVSqx7WAKbvl7KsIjx/Lp5rCBvMlP785+OD3xOXLkiFNf3kxZWRmvcm/BU/0Uol2+dbqi56JxNpa8yULe5Ke3p3li40wAwDc5DciubcNn4loAwNpYH7x311UYFeCLwoZ2PPbvbJgt7LZwObVq66rQVWO6AQCL466wlp0va2Hba9biqWO56Daaoe02orixHYxrW8sNecib/PTu7IfLGxh6I442QDIajQNuojVQuTdtxOXofRhO7fKt0xU9F42zseRNFvImP31/mt1HMnCuTAGRCGAYwN9XhKSn12LcFaF4+fsSfNjr3K8pkaMgU7L3+cRFjsLpJ1YhICAADeourHr1LPx8RJgRE2Zzb1BooB8MJgsMZgseu2Y6nrp+ljUf8ubQa3coebMHQY+sGMmkpaXxKvcWPNVPIdrlW6crei4aZ2PJmyzkTX76/jRv3LYQ2xbGoufP5E1zx6O8gN35+YE1UxF4cUfmCWOC8d5dV1l1ty2ZhPT0dABAbHgQYkYHwmRhrJOe5zfPRmSQCB16EwxmdmuUd89KcSK3gXPuQx3yJj+9O/tBKz694DJj5Io3/eVCeBfkTcJZxFVK/FzSjAfXTMPY0Zee8n31h1K8f6ESL22fi3uvjsPT/8lDdk0bvvy/FYgOu/RAzMOfZuHHIjkAYGxYIMT7rgXDAEWN7QgO8MV/surwwYUqBPr54Nj/rUDCuJB+vdnS3o0xo+wfg0GMPGjFRyB6Nk9ytdxb8FQ/hWiXb52u6LlonI0lb7KQN/npHWmWT4vE725KwNjRQTaxz94wC2eeWod7lrPnfB3YsQBnn16P6LBAm7ir4sZYv186NQIikQjJyUmYNzEc08eG4tkbZuOa2WOhN1nw0KcSKLR6u3mcLm7GspfPYOlffsazX+Uhq6aNc18HG/ImP707+0ETHw5cddVVvMq9BU/1U4h2+dbpip6LxtlY8iYLeZOf3lVv+viIEB8davcQyd5xiyZfYf1+6cWbnXuX+/qI8PYdCzFjbCia2/V473yl3bbPljYDADRdRhyT1GPH+6k2k5/zZS1IrWx1ui+DAXmTn96d/aCJDwfoyRkWejqBn56e6hIO8iY/vdDenDchHAG+7K+dpVMj7NYTFuSPl7bPBQD8J6sBajuLPsWN7D1Cv1ofj9XTo2BhgBf/VwSLhcH7Fyqx60gm7vqHGJ+myWx0TZoubH03Gccy65zK3Z2QN/np3dkPmvhwgPZKYaH9KPjpaR8f4SBv8tML7c0gf18cuG0Bfn/THFwZG95vPVdPi8SyqREwmhmcabT9NWUyW1Aq1wIAdiyeiDdvX4iwQD8UNGjw0nfFePWHS2c2/eF/Rai9+IQZAHyRUYeCBg2e/W8+lB32L6MJBXmTn94j+/gMdWQyGfbs2YOpU6ciODgY8fHx2L9/PwwGg9vaoN1xWWgHUn562rlZOMib/PSD4c2bF8TigTXTHNbzyIbpAICsVpHN3j4yZSf0JguC/X0RFxmC6LBA7F49FQBwJEUGANgyfzwWTR4DhgEyZCqrVtN16YilDxKr0KLtxp6jmfhncrVTfeEDeZOf3iM7Nw91SktLYbFY8MEHH6CoqAhvvvkm3n//fezbt89tbdB5SCx05gw/PZ3VJRzkTX76oeTNFdMiEejng06TCJUK9pee2cKg6OJlrtnjw+Drw95TdOtVE220ty2ZhCXWDRMv3ftT39Zl/f6rrHrc+HYyzpS24E8niwXfOJG8yU/vkbO6hjqbNm3Cpk2brD9PmzYNZWVlOHToEA4cOOCWNq644gpe5d6Cp/opRLt863RFz0XjbCx5k4W8yU8/lLwZ4OeDhZPCIa5uQ3q1Cj+VKHA0VQa1jl21SRh/6ZHlyZGjsHTKFciUtSEqNAAr4yOh7TYBqEZundoa13OAKgCoOm2vBig7DYgKDYRQkDf56d3ZD6+Z+NhDo9EgIiKi33K9Xg+9/tJ13vZ29i8Jo9Fo99T56urqAesbqLynPm84zd7R+zCc2uVbpyt6LhpnY8mbLORNfvqh5s2rJo6GuLoNfzxZ2qdsVkyIjWfvXT4JmbI23LFkIhiLGfNiQwGwewS1tuswOsgPtSp25SgmLBDNlz0qX9qoxvKpwnmHvMlP70jDZfzy2g0MpVIpFi9ejAMHDuDBBx+0G/Piiy/ij3/8Y5/X//3vf2PUqFFCp0gQBEEMQKlahEMlvtafb5liRnOXCDUdIjw024zwANt4jQEI8wd8Lh6t8YcsX7QbL14Om2rGV9W+EIHB6hgGSc3snR4xwQyau0S4daoZCWMYnG/ywdwIBrPC3ferkWEACwDfvk/7E25Cp9PhrrvucmoDwyE/8Xnuuefw2muvDRhTUlKC2bNnW39uaGjAunXrsH79ehw+fLhfnb0Vn0mTJqG1tdXuG5eamoqVK1f2W99A5UajEadPn8Z111037HfHdfQ+DKd2+dbpip6LxtlY8iYLeZOffqh5U93ZhaWvJgEA4iJG4fSTq+zuFdQfb/4sxd8vVNm8FhsehA/uWYRfHErH8qkRSBgfhn8ky3DdnLE4XcIenjo3djS++dXVTrfjiPuOSFDe1IYzT2/AqIDBvdAyUrzZ3t6OqKgo75j4KBQKKJXKAWOmTZuGgAB26t/Y2Ij169fj6quvxtGjR+Hj4/z923RkBTESIW8SQxWj0Yib3/gRUq0vvvy/FbhqMvf7PPQmMx74WIKkCnZDw/kTw/G/R1ejTqXDFSEB+KlQjqf+k2ej8fcVoeIvN7qlDx16E+bu/wkA8Mn9y7B2ZrRb6iVs8aojK6KjozF79uwBv3omPQ0NDVi/fj0WL16MI0eOcJr0OAMdC8BCW6/z09ORFcJB3uSnH4revH+mBWf3rnFp0gMAgX6+OHTPYqyaHgkAWDqFvU9kUsQohAb6Yda4sD4ao5mBpssIs4VBTm0bLBbX1wdkrZcew+7vCA4hGYnedMSQX/Fxlp5JT1xcHD7++GP4+l66Ljxu3Din6nA0Y+zu7kZQUJAdpeNyb/qr2tH7MJza5VunK3ouGmdjyZss5E1+em/2Zs8kZu6EcAT5X/r9YDJbsOil0xefArvEycdW4/OMWnwmrsULWxKw5+JeQVz5Nq8Rj32eAwB47JrpeOr6Wa53wgVGije9asXHWU6fPg2pVIozZ85g4sSJGD9+vPXLXeTl5fEq9xY81U8h2uVbpyt6LhpnY8mbLORNfnpv9qavjwhLpkTYTHoAwM/XB8umXHpaKMif/bX4Y6Ecn4lrAQDvnKkAwE6e3jlTgWOZdTicxG6AeDlyTTe2HUzBMQl7LEbvFZ/q1sHfxHAketMRXjPx2bVrFxiGsfvlLqZOHXjG76jcW/BUP4Vol2+drui5aJyNJW+ykDf56UeqN+MiQ6zfr585FgDw3jmp9TVNlxGVig58m9eIv50ux7P/zcefvyvBg59kwXzZZbB3z1Ygr06NZ7/KB2A72ZEpB3/iMxK96QivmfgMBo5usnZU7i14qp9CtMu3Tlf0XDTOxpI3Wcib/PQj1Zu7V01BaKAfbrgyBlOiQuzGnC1pwYeJtk+I5dWpcTK/0ea13pfMLBYGVR6+x2cketMRNPHhgKNrzMP9/ghn8VQ/hWiXb52u6LlonI0lb7KQN/npR6o3J0WMQtrz1+DdO6/CpIhgm7LHr50BADhd0ozipvY+2vSqS7+QTWYL/Hpt1vP6T2U2O0erOg2CH41xOSPRm46giQ8HgoODeZV7C57qpxDt8q3TFT0XjbOx5E0W8iY//Uj2ZliQPwL8fLDpynEIC2L32tkwKxor49mnwTIvHnY6NiwQ3z66Gm/sWAAAyK5RAwD+mVyNuS/+hO8Lmqx1vn+h0qYNo5mBVm97E3Vv3H17BjAyvekImvhwQC6X8yr3FjzVTyHa5VunK3ouGmdjyZss5E1+evImEBkaiAvPbMDzm2fjxZuvxPyJ4fD1EaFnPjLhimDMmxiONTOjAADlLVq0dxvxp5PF6DZa0G202K038OJKkKrD0KeMYRhodEZs/3sqbnonGSaz/TpcYSR60xE08eHArFkDP4boqNxb8FQ/hWiXb52u6LlonI0lb7KQN/npyZssESEB+L918YiLDMGoAD+bA1EnjGFXHsaGBWFyxCgwDJAqbe1Tx7SoEGusr48IkaHsfnNfZ9fDZLag22iGyWzB6eJmLPnzz9h2MBl5dWoUN7WjQd3Vp75XfyjF3YfTYTBxmxSNRG86giY+HMjOzuZV7i14qp9CtMu3Tlf0XDTOxpI3Wcib/PTkTftcNXmM9fuJV1w6x3HzPHaPuE/Savpo/vfYapx8bDVuWTQBnz2wHIFgD9F856wU17+ZiIV/OoUdH6ThwU8kUHYaIFNeOjl+3V/PQ6659Ki8yWzB+xcqkSJVIqNaxSn3kehNR3jNBobugI6sIEYi5E1iqDJUvHkitwFPfJELAHhp+1zce3UcAEDaosXGv/XdUTg82B95+6+3eW3z20kosXNzdH/cuWwy1s2Mwg+Fcjywehq2vpcMAPh0zzKsmUHHXlzOiNzAcDCgYwFY6FgAfvqheCyAt0De5Kcnb9qn93EZE8dcusl2+tgwTIkcZRMrEgHv3rmoTx3SZucnPQDQqO7Cw//KxoncRvzueIH19fau/m+OtsdI9KYjaOLDAUenyXriVGhP4Kl+CtEu3zpd0XPROBtL3mQhb/LTkzftM/GKYEyJHAU/HxFmxITalC2fGmn9ftOV4/Dz3nV2DyLdPI/bKQLt3Ubr9/n1Guv3mi729cRyBTb+7YLDS18j0ZuOoIkPgIMHDyIhIQFLly4FAKhUKqSkpMBkMllnmYmJiUhMTER2djYaGhoglUpRUlIChUIBsVgMvV6Pzz//3Brb1dWFjIwMNDc3o6ysDFKpFDqdDtnZ2ejs7LSp12g0IjU1Fa2trSgqKkJVVRXq6uqQm5sLjUZjE2uxWJCUlAS1Wo28vDzU1tZCJpOhsLAQSqXSbt4dHR3IysqyybulpQVisRjd3d02sd3d3da8S0tLUVFRgcbGRmRlZaGjowOJiYkQi8U2eSuVShQWFqK6uhq1tbXIy8uz5s0wjM2/Go3Gmnd1dbU179TUVBiNRrt5NzY2oqKiAt9++y2am5uRkZFhN2+xWIyWlhaUlJRAKpWioaHBJu+eWJPJhJSUFCiVSpw4cQIymcyat1qtRlJSEiwWi41Go9EgNzcXdXV1qKqqQlFREVpbW3Hs2LE+eXd2dkIikaCpqQnl5eUoKyuz5t3V1WXjE71eD7FYDIVCYZN3dnY2tFqtTWxP3iqVCgUFBZDJZKipqUF+fj7Onz+P5OTkPnm3t7cjJycHp06dQmVlpTXvtLQ0GAwGa6xcLodOp4NEIoFcLrfmLZfLkZmZCZ1OZ1OvXq9Heno6FAoFiouLUVlZifr6euTk5ECr1drEms1mJCcno62tDfn5+aipqbHm3dbWhuTkZJjNZhuNVqtFTk4O6uvrUVlZieLiYigUCqSnp0Ov19vE6nQ6ZGZmQi6X4+TJkygvL4dcLodEIumTt8FgQFpamvWz1jvv9vb2Pp+15ORkqNVqa94ymQwFBQV9xoiez4RWqx1wjOhd/+VjRHl5OZqamiCRSNDZ2Wnzf+/KGHH+/HnOY8SpU6ecHiPOnDnT7xjRO+9jx44NOEakp6f3O0bIZDK0tbVBpVI5NUaUlpa6fYwoLCxETU0NXt08Ca9sjEaoyGAzRiyOu7QadMNECwK622zGiJ68E7qLsXSK8wet5tSq7b4uLqtDeXUtdv4zA9KWDtz9jzToTWZr3mcvJNuMET/++CPy8/OhVqsHHCN6Pmv9jRE9nzWJRIKff/6Z1xjR8/uUyxiRlpY24BhRWlrq9HtL9/j0wtE1QqVSicjISDtKx+VD5Vq1O3D0PgyndvnW6Yqei8bZWPImC3mTn568yR1NlxE7PxJj2dQI/O6mhH7jlEolmvX+uPGdJJvX/7x9Ln5/vJBXDrtXTcH+rVciVdqKe/+ZgWdumIWH18Vb2x0J3qR7fASiqamJV7m34Kl+CtEu3zpd0XPROBtL3mQhb/LTkze5Ex7sjxOPrh5w0gOw/UyIHY1Xb5mHT+5fhn03zsbbdyzEPVfH4cWtCdg8dxxO/2YtHtkQzzmHIykyAMCfThbDbGHw6g+XVj8aGxv7UbnOUPemI/zcVtMIIDQ0lFe5t+CpfgrRLt86XdFz0TgbS95kIW/y05M3haOnn3csmwwANvcB7Vo1FbtWsYdwPnPDbFwoV6CwgdvN0AzDIPCyk+c79SY8/K0c0zJ0+HTPcj7p2zDUvekIWvHhgI/PwG+Xo3JvwVP9FKJdvnW6oueicTaWvMlC3uSnJ28KB5d+Tuj15NiKac5dEnrnjBSjgy6tZWi7jThVLEdDuwFJFa3oNpoBAGdKmnGurMXpXOwx1L3psC631TQCaGtr41XuLXiqn0K0y7dOV/RcNM7GkjdZyJv89ORN4eDSz/joS6sbL2xJQNKzG5D07Aa8uDUBF55ZD/9eB6H28ObP5UiquLSD9OcZtfjtfy89Bi/XdEOjM2LPxxLsPpJpnQi5wlD3piPoUhcH4uLieJV7C57qpxDt8q3TFT0XjbOx5E0W8iY/PXlTOLj081fr42E0W2CyMJgzPgwiETvR6bkc9s2vV+FwUhWe2TQb5c1a7D6S2aeOl7+3fcopsUKB5vZLu0G3duhtdqHmwlD3piNoxYcDBQUFvMq9BU/1U4h2+dbpip6LxtlY8iYLeZOfnrwpHFz6GRbkj9/dlID9W6+0Tnp6M3dCON66YxEmjAlGbLhzp5b/4UQRDp67dFq88rLDUgsbNHjr53KnVoKGujcdQY+z98LR43AWi2XA64wDlQ+HxzKdxdH7MJza5VunK3ouGmdjyZss5E1+evKmcAjlzfZuI+a/eIqz7qP7liDQzxf3fCRGoJ8P9BcPP31+82z83zr7T5bVqXQYHx4EHxG/e26E8CY9zi4QycnJvMq9BU/1U4h2+dbpip6LxtlY8iYLeZOfnrwpHEL1c3RQ3wnhoxumIzzYH4fuvgrbpgfY1f39fCXu+UgMANZJDwBUKTrtxidVKLDm9XPY/78iu33pNprR2qF3KmehvekIWvHpBR1SSoxEyJvEUIW86RxTnvvO+v3j187A3utmWn/+PKMWz3/NXiZaOuUKZMoGvkn44XXxeG7z7D6v3/h2EoovHrIqe/WmPuV3/SMdqZVKJD6zAZMjXbt3iA+04iMQ3nDYnjuggyD56ekgSOEgb/LTkzeFQ8h+vrR9LgDgjqWT8JuNM2zKaqsqrN8f3rkUb9+xcMC6RgdfeuapvduIFGkrLBYGXb3u/Tn437M4mlKN3usmqZVKAMAXmbUO8/X0IaX0VBcHFixYwKvcW/BUP4Vol2+drui5aJyNJW+ykDf56cmbwiFkP++9Og5XTR6D+OjQPjdD37pmPv5ZKMHcCeEIH+WPzXPH4wnk9luX4eJlL4Zh8NAnEqRXqbB8agTqVDprzF8zuwAUIzTIH7cunmgzAWpQdznMV2hvOoJWfDhQXV3Nq9xb8FQ/hWiXb52u6LlonI0lb7KQN/npyZvCIXQ/r4wNR9BlOzcDgFbRgNTnrsG/H2R3bg7w87F+b48TuY2QyFR49qt8pFexJ7+Lq1UwWfreFXOmpBlyTTfkvR6Tb+8y9om7HKG96Qha8eGAo0PVPHE4oifwVD+FaJdvna7ouWicjSVvspA3+enJm8LhSW9GhgbavDYzJqzf+OrWTtz6fppTddcoddhw4DzMvVZ86tq6kFzRiqunRcDP1/7aitDedASt+HDAaBx4Juuo3FvwVD+FaJdvna7ouWicjSVvspA3+enJm8IxlLwZ1WsiFOjng29+vdKlw1GLm9rRZTRbL48BgLSlA/d8JMZXWfWccnKEO98/mvgAOHjwIBISErB06VIAgEqlQkpKCkwmk/WGqsTERKhUKmRnZ6OhoQFSqRQlJSVQKBQQi8XQ6/XIzs62xnZ1dSEjIwPNzc0oKyuDVCqFTqdDdnY2Ojs7beo1Go1ITU1Fa2srioqKUFVVhbq6OuTm5kKj0djEWiwWJCUlQa1WIy8vD7W1tZDJZCgsLIRSqbSbd0dHB7KysmzybmlpgVgsRnd3t01sd3e3Ne/S0lJUVFSgsbERWVlZ6OjosPatd95KpRKFhYWorq5GbW0t8vLyrHkzDGPzr0ajseZdXV1tzTs1NRVGo9Fu3o2NjaioqIBUKkVzczMyMjLs5i0Wi9HS0oKSkhJIpVI0NDTY5N0TazKZkJKSAqVSifLycshkMmvearUaSUlJsFgsNhqNRoPc3FzU1dWhqqoKRUVFaG1tRV5eXp+8Ozs7IZFI0NTUhPLycpSVlVnz7urqsvGJXq+HWCyGQqGwyTs7OxtardYmtidvlUqFgoICyGQy1NTUID8/H62trUhOTu6Td3t7O3JyclBfX4/Kykpr3mlpaTAYDNZYuVwOnU4HiUQCuVxuzVsulyMzMxM6nc6mXr1ej/T0dCgUChQXF6OyshL19fXIycmBVqu1iTWbzUhOTkZbWxvy8/NRU1NjzbutrQ3Jyckwm802Gq1Wa5N3cXExFAoF0tPTodfrbWJ1Oh0yMzMhl8tRWVmJ8vJyyOVySCSSPnkbDAakpaVZP2u9825vb+/zWUtOToZarbbmLZPJUFBQ0GeM6PlM9Pyf9TdG9K7/8jGivLwcTU1NkEgk6OzstPm/d2WM6PGnUGOEXC7vd4zonXdeXt6AY0TP/5G9MUImk6GtrQ0qlcqpMaK0tNTtY0RhYSGvMaIn797/n47GiMs/awONEfby7j1G1NbWIj8/H2q1us8YAVx8lF0pg8jg+N4cLnyTU28zRhSWVqCgvBo5OTlQqVScx4jLf29ePkaUlpYOlI4N9Dh7Lxw9Dtfa2oqoqKh+9QOVe9NjmY7eh+HULt86XdFz0TgbS95kIW/y05M3hWOoebP3I/CyV2/Cp2kyvHCiyG4dpS9twmfiWvxQ0ARJjXNnZt29fDJe2JIAbbcJ0WGBuOPDNEhkbfh57zqEosvt3qTH2QWivLycV7m34Kl+CtEu3zpd0XPROBtL3mQhb/LTkzeFY6h5863bFwIAnt00CwB7b05/BPn7Ys/qqfjbbQudbtff1wcPfiLBspd/RnVrJ9Kr2Buk/5lSLbg3HUErPr1wNGM0Go0D/tUxULk3/eXi6H0YTu3yrdMVPReNs7HkTRbyJj89eVM4hqI3FVo9okIDIBKJkFiuwM5/ZtiN69mwkGEYxO/7HnYe8BqQWxdPtLnn58xvViM+JpxTHY7eP1rxEYi0tIHvdHdU7i14qp9CtMu3Tlf0XDTOxpI3Wcib/PTkTeEYit6MDgu07vuzZkYUvnp4BeaMt500vHvnIuv3IpEIqc9di1fX9t2Z+bVfzuu3nUpFh83PL30ldir33rjz/aMVn17QkRXESIS8SQxVyJuDz50fpiOtit2F+dM9y7BmRrTduKtfPgN5ezcWTR6DLx9agQA/Hxw8J8Vffypz2MYtiybgbxcvtbkLWvERCNp6nYWOBeCnp2MBhIO8yU9P3hSO4eLN3hs/9zfpSUxMxOm9a3HhmfX45terEODHTiV+tS4eT152ZIY9WlqaOeXU06a7oA0MObB48WJe5d6Cp/opRLt863RFz0XjbCx5k4W8yU9P3hQOb/NmSJA/wi47Gd7HR4QlcREO9SGjr3CpTXdBKz4cKCkp4VXuLXiqn0K0y7dOV/RcNM7GkjdZyJv89ORN4Rgu3rzsqC/OdcaOCXKoV7RpuKTksE2u0MSHAxMmTOBV7i14qp9CtMu3Tlf0XDTOxpI3Wcib/PTkTeEYLt4UwfHMZ6A6Y8cEW7/ftXKK/SC/AE45OWqTKzTx4YBWq+VV7i14qp9CtMu3Tlf0XDTOxpI3Wcib/PTkTeEYKd4M8vfF2DD2SIyrp0Xi7FPrsGZGFCZFBOOeqycDADq7uR8/4c5+0D0+HHD0ANxIeUDOU/0Uol2+dbqi56JxNpa8yULe5KcnbwrHcPHm+HDHl6oc1XnqN2uRUa3CdQkxEIlE+HQPexp8qrQV/0qvRbdJWG86giY+HBgzZgyvcm/BU/0Uol2+dbqi56JxNpa8yULe5KcnbwrHcPHmc5tnQ91lxB1LJ7lc55hRAbj+ynF9Xg8K8AUAGC19ihzizvePLnVxoKamhle5t+CpfgrRLt86XdFz0TgbS95kIW/y05M3hWO4eDMyNBD/2LkE186JcVudPQT7sxOfTj33S13ufP9oxYcD8+b1vzOlM+Xegqf6KUS7fOt0Rc9F42wseZOFvMlPT94UDvIme+PzgR0LEOjDfcnHnf2gFR8OZGZm8ir3FjzVTyHa5VunK3ouGmdjyZss5E1+evKmcJA3gfBgf9y6eCLCtbJBa9MeNPEBcPDgQSQkJGDp0qUAAJVKhZSUFJhMJutukYmJiVi0aBGys7PR0NAAqVSKkpISKBQKiMVi6PV6a32JiYno6upCRkYGmpubUVZWBqlUCp1Oh+zsbHR2dtrUazQakZqaitbWVhQVFaGqqgp1dXXIzc2FRqOxibVYLEhKSoJarUZeXh5qa2shk8lQWFgIpVJpN++Ojg5kZWXZ5N3S0gKxWIzu7m6b2O7ubmvepaWlqKioQGNjI7KystDR0YHExESsXbvWJm+lUonCwkJUV1ejtrYWeXl51rwZhrH5V6PRWPOurq625p2amgqj0Wg378bGRlRUVGDs2LFobm5GRkaG3bzFYjFaWlpQUlICqVSKhoYGm7x7Yk0mE1JSUqBUKhEREQGZTGbNW61WIykpCRaLxUaj0WiQm5uLuro6VFVVoaioCK2trfDz8+uTd2dnJyQSCZqamlBeXo6ysjJr3l1dl05ATkxMhF6vh1gshkKhsMk7Ozvb5imG3nmrVCoUFBRAJpOhpqYG+fn5mD9/PpKTk/vk3d7ejpycHEybNg2VlZXWvNPS0mAwGKyxcrkcOp0OEokEcrncmrdcLkdmZiZ0Op1NvXq9Hunp6VAoFCguLkZlZSXq6+uRk5MDrVZrE2s2m5GcnIy2tjbk5+ejpqbGmndbWxuSk5NhNpttNFqtFjk5Oaivr0dlZSWKi4uhUCiQnp4OvV5vE6vT6ZCZmQm5XI6YmBiUl5dDLpdDIpH0ydtgMCAtLc36Weudd3t7e5/PWnJyMtRqtTVvmUyGgoKCPmNEz2dCq9UOOEb0rv/yMaK8vBxNTU2QSCTo7Oy0+b93ZYyYP38+5zEiPj7e6TFi1qxZ/Y4RvfP28/MbcIxYs2ZNv2OETCZDW1sbVCqVU2NEaWmp28eIwsJCXmNET969/z8djRGXf9YGGiPs5d17jIiLi0N+fj7UavWAY0TPZ62/MaLnsyaRSDBz5kxeY8SiRYs4jxGrVq0acIwoLS2F0zCEFY1GwwBgNBqN3fILFy4MqB+o3GAwMMePH2cMBgOvHIcCjt6H4dQu3zpd0XPROBtL3mQhb/LTkzeFg7zJT+9I4+j3d2/okNJeODrkTK/XIzAwsF/9QOXedNieo/dhOLXLt05X9Fw0zsaSN1nIm/z05E3hIG/y0zvS0CGlApGbm8ur3FvwVD+FaJdvna7ouWicjSVvspA3+enJm8JB3uSnd2c/aOLDgWnTpvEq9xY81U8h2uVbpyt6LhpnY8mbLORNfnrypnCQN/np3dkPmvhwoLW1lVe5t+CpfgrRLt86XdFz0TgbS95kIW/y05M3hYO8yU/vzn7QxIcDjq4xD/dr0M7iqX4K0S7fOl3Rc9E4G0veZCFv8tOTN4WDvMlP785+0MSHA8HBwbzKvQVP9VOIdvnW6Yqei8bZWPImC3mTn568KRzkTX56d/aDJj4caG5u5lXuLXiqn0K0y7dOV/RcNM7GkjdZyJv89ORN4SBv8tO7sx808eHAjBkzeJV7C57qpxDt8q3TFT0XjbOx5E0W8iY/PXlTOMib/PTu7AdNfDiQk5PDq9xb8FQ/hWiXb52u6LlonI0lb7KQN/npyZvCQd7kp3dnP2gDw15w2QCJK960ERfhXZA3iaEKeZNwFq/bwFAmk2HPnj2YOnUqgoODER8fj/3798NgMFhjzp8/j23btmH8+PEICQnBwoUL8dlnn7k1j55zQlwt9xY81U8h2uVbpyt6LhpnY8mbLORNfnrypnCQN/np3dmPYbHi8+OPP+LLL7/EnXfeienTp6OwsBAPPvgg7r33Xhw4cAAA8PLLL6OrqwubN29GTEwMTp48ib179+LEiRPYsmWLU+04mjGaTCb4+fn1qx+o3Jv+cnH0PgyndvnW6Yqei8bZWPImC3mTn568KRzkTX56RxqvW/HZtGkTjhw5guuvvx7Tpk3DzTffjKeffhpff/21NWbfvn146aWXsHLlSsTHx+OJJ57Apk2bbGL4IhaLeZV7C57qpxDt8q3TFT0XjbOx5E0W8iY/PXlTOMib/PTu7MfgTz/dhEajQUREhMOYOXPm9Fuu1+uh1+tt4gFApVLBaDT2iR83bhyUSmW/9Q1UbjQaodPpoFQqh/1fLo7eh+HULt86XdFz0TgbS95kIW/y05M3hYO8yU/vSKPVagEATl3E4nw2/BCgoqKCGT16NPPhhx/2G/Pll18yAQEBTGFhYb8x+/fvZwDQF33RF33RF33Rlxd81dXVOZxDePQen+eeew6vvfbagDElJSWYPXu29eeGhgasW7cO69evx+HDh+1qzp07hy1btuDQoUPYuXNnv3VfvuJjsVigUqkQGRkJkUjUJ37p0qXIzMzst76Bytvb2zFp0iTU1dW5/YmxwcbR+zCc2uVbpyt6LhpnY8mbLORNfnrypnCQN/npHWkYhoFWq0VsbCx8fAa+i8ejl7qeeuop7Nq1a8CY3ieyNjY2YsOGDVi5ciU+/PBDu/EXLlzA1q1b8eabbw446QGAwMBABAYG2rw2ZsyYfuN9fX0H/PA5KgeA0aNHD/sPsDP9HC7t8q3TFT0XjbOx5E0W8iY/PXlTOMib/PTOaMLDw52qy6MTn+joaERHRzsV29DQgA0bNmDx4sU4cuSI3Rnd+fPnsWXLFrz22mt46KGH3J0uHnnkEV7l3oKn+ilEu3zrdEXPReNsLHmThbzJT0/eFA7yJj+9O/sxLB5nb2howPr16xEXF4ePP/4Yvr6+1rJx48YBuHR564knnsDjjz9uLQ8ICHB4E/RgIOTmiATBB/ImMVQhbxJCMCye6jp9+jSkUimkUikmTpxoU9Yzb/v444+h0+nwyiuv4JVXXrGWr1u3DufPnx/MdO0SGBiI/fv397m0RhCehrxJDFXIm4QQDIsVH4IgCIIgCHcwLDYwJAiCIAiCcAc08SEIgiAIYsRAEx+CIAiCIEYMNPEhCIIgCGLEQBMfgiAIgiBGDDTxGQKcPHkSs2bNwowZM/o9hoMgPMEvfvELXHHFFbj11ls9nQpB2FBXV4f169cjISEB8+fPx3/+8x9Pp0QME+hxdg9jMpmQkJCAc+fOITw8HIsXL0ZqaioiIyM9nRpB4Pz589Bqtfj444/x1VdfeTodgrDS1NSE5uZmLFy4EHK5HIsXL0Z5eTlCQkI8nRoxxKEVHw+TkZGBK6+8EhMmTEBoaCg2b96MU6dOeTotggAArF+/HmFhYZ5OgyD6MH78eCxcuBAAu4N/VFQUVCqVZ5MihgU08eFJYmIitm7ditjYWIhEIhw/frxPzMGDBzFlyhQEBQVh+fLlyMjIsJY1NjZiwoQJ1p8nTJiAhoaGwUid8HL4epMghMSd/szKyoLZbMakSZMEzprwBmjiw5POzk4sWLAABw8etFv+5ZdfYu/evdi/fz+ys7OxYMEC3HDDDWhpaRnkTImRBnmTGMq4y58qlQo7d+7Ehx9+OBhpE94AQ7gNAMw333xj89qyZcuYRx55xPqz2WxmYmNjmVdeeYVhGIZJSUlhtm/fbi1/4oknmM8++2xQ8iVGDq54s4dz584xv/zlLwcjTWKE4qo/u7u7mTVr1jCffPLJYKVKeAG04iMgBoMBWVlZ2Lhxo/U1Hx8fbNy4EWlpaQCAZcuWobCwEA0NDejo6MAPP/yAG264wVMpEyMEZ7xJEJ7CGX8yDINdu3bhmmuuwb333uupVIlhCE18BKS1tRVmsxkxMTE2r8fExEAulwMA/Pz88MYbb2DDhg1YuHAhnnrqKXqiixAcZ7wJABs3bsSOHTvw/fffY+LEiTQpIgYFZ/yZkpKCL7/8EsePH8fChQuxcOFCFBQUeCJdYpjh5+kECODmm2/GzTff7Ok0CKIPP//8s6dTIAi7rF69GhaLxdNpEMMQWvERkKioKPj6+qK5udnm9ebmZowbN85DWREEeZMY2pA/CSGhiY+ABAQEYPHixThz5oz1NYvFgjNnzmDFihUezIwY6ZA3iaEM+ZMQErrUxZOOjg5IpVLrz9XV1cjNzUVERAQmT56MvXv34r777sOSJUuwbNkyvPXWW+js7MTu3bs9mDUxEiBvEkMZ8ifhMTz9WNlw59y5cwyAPl/33XefNebdd99lJk+ezAQEBDDLli1j0tPTPZcwMWIgbxJDGfIn4SnorC6CIAiCIEYMdI8PQRAEQRAjBpr4EARBEAQxYqCJD0EQBEEQIwaa+BAEQRAEMWKgiQ9BEARBECMGmvgQBEEQBDFioIkPQRAEQRAjBpr4EARBEAQxYqCJD0EQBEEQIwaa+BAEITi7du2CSCTq89X7rCaCIIjBgA4pJQhiUNi0aROOHDli81p0dLTNzwaDAQEBAYOZFkEQIwxa8SEIYlAIDAzEuHHjbL6uvfZaPProo3jyyScRFRWFG264AQBQWFiIzZs3IzQ0FDExMbj33nvR2tpqrauzsxM7d+5EaGgoxo8fjzfeeAPr16/Hk08+aY0RiUQ4fvy4TQ5jxozB0aNHrT/X1dXhtttuw5gxYxAREYFt27ZBJpNZy3ft2oXt27fjwIEDGD9+PCIjI/HII4/AaDRaY/R6PX77299i0qRJCAwMxPTp0/HRRx+BYRhMnz4dBw4csMkhNzeXVrsIwoPQxIcgCI/y8ccfIyAgACkpKXj//fehVqtxzTXXYNGiRZBIJPjxxx/R3NyM2267zap55plncOHCBZw4cQKnTp3C+fPnkZ2dzaldo9GIG264AWFhYUhKSkJKSgpCQ0OxadMmGAwGa9y5c+dQWVmJc+fO4eOPP8bRo0dtJk87d+7E559/jnfeeQclJSX44IMPEBoaCpFIhPvvv7/PKteRI0ewdu1aTJ8+3bU3jCAIfnj4dHiCIEYA9913H+Pr68uEhIRYv2699VZm3bp1zKJFi2xiX3rpJeb666+3ea2uro4BwJSVlTFarZYJCAhgjh07Zi1XKpVMcHAw88QTT1hfA8B88803NvWEh4czR44cYRiGYT799FNm1qxZjMVisZbr9XomODiY+emnn6x5x8XFMSaTyRqzY8cO5vbbb2cYhmHKysoYAMzp06ft9ruhoYHx9fVlxGIxwzAMYzAYmKioKObo0aNOvGsEQQgB3eNDEMSgsGHDBhw6dMj6c0hICO68804sXrzYJi4vLw/nzp1DaGhonzoqKyvR1dUFg8GA5cuXW1+PiIjArFmzOOWTl5cHqVSKsLAwm9e7u7tRWVlp/fnKK6+Er6+v9efx48ejoKAAAHvZytfXF+vWrbPbRmxsLG666Sb885//xLJly/Dtt99Cr9djx44dnHIlCMJ90MSHIIhBISQkxO7lnZCQEJufOzo6sHXrVrz22mt9YsePH+/0vTEikQgMw9i81vvenI6ODixevBifffZZH23vm679/f371GuxWAAAwcHBDvN44IEHcO+99+LNN9/EkSNHcPvtt2PUqFFO9YEgCPdDEx+CIIYUV111Ff773/9iypQp8PPrO0TFx8fD398fYrEYkydPBgC0tbWhvLzcZuUlOjoaTU1N1p8rKiqg0+ls2vnyyy8xduxYjB492qVc582bB4vFggsXLmDjxo12Y2688UaEhITg0KFD+PHHH5GYmOhSWwRBuAe6uZkgiCHFI488ApVKhTvvvBOZmZmorKzETz/9hN27d8NsNiM0NBR79uzBM888g7Nnz6KwsBC7du2Cj4/tcHbNNdfgvffeQ05ODiQSCR5++GGb1Zu7774bUVFR2LZtG5KSklBdXY3z58/j8ccfR319vVO5TpkyBffddx/uv/9+HD9+3FrHsWPHrDG+vr7YtWsXnn/+ecyYMQMrVqxwzxtFEIRL0MSHIIghRWxsLFJSUmA2m3H99ddj3rx5ePLJJzFmzBjr5Oavf/0r1qxZg61bt2Ljxo1YvXp1n3uF3njjDUyaNAlr1qzBXXfdhaefftrmEtOoUaOQmJiIyZMn45ZbbsGcOXOwZ88edHd3c1oBOnToEG699Vb8+te/xuzZs/Hggw+is7PTJmbPnj0wGAzYvXs3j3eGIAh3IGIuvwhOEAQxDFm/fj0WLlyIt956y9Op9CEpKQnXXnst6urqEBMT4+l0CGJEQ/f4EARBCIRer4dCocCLL76IHTt20KSHIIYAdKmLIAhCID7//HPExcVBrVbj9ddf93Q6BEGALnURBEEQBDGCoBUfgiAIgiBGDDTxIQiCIAhixEATH4IgCIIgRgw08SEIgiAIYsRAEx+CIAiCIEYMNPEhCIIgCGLEQBMfgiAIgiBGDDTxIQiCIAhixEATH4IgCIIgRgz/D9PFFRuEYWL0AAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## compare power spectral density of closing prices with Brownian noise\n",
    "df = pd.read_csv('data/OHLC_1h.csv', header=0, sep=',')\n",
    "ng = NoiseGenerator()\n",
    "brown_noise = ng.generate(dt=1e-5, n=10000, colour=ng.brown()) * 1000\n",
    "\n",
    "# draw graph\n",
    "plt.psd(brown_noise, 512, 1 / 0.001, label='brown noise')\n",
    "plt.psd(df['close'].to_numpy(), 1000, 1 / 0.001, label='close_price_1h')\n",
    "plt.xscale('log',base=10)\n",
    "plt.minorticks_on()\n",
    "plt.grid(which='minor', color='grey', linestyle=':', linewidth=0.5)\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The Power Spectral Density graph is as expected, and tells us that:\n",
    "1. There are no dominant (long-term) cyclical pattens in the data.\n",
    "2. There is some structure to the data... A 6dB/octave roll off. (NB: This 'structure' is also shared by brown noise and random walks).\n",
    "4. First-order filters (which are characterised by a similar roll off) are therefore unable to fully detrend the data, so should be used/interpreted with care.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hurst exponent with 20 lags: 0.496 \t(brown noise = 0.517)\n",
      "Hurst exponent with 100 lags: 0.487 \t(brown noise = 0.509)\n",
      "Hurst exponent with 300 lags: 0.504 \t(brown noise = 0.496)\n",
      "Hurst exponent with 500 lags: 0.509 \t(brown noise = 0.506)\n",
      "Hurst exponent with 1000 lags: 0.517 \t(brown noise = 0.505)\n"
     ]
    }
   ],
   "source": [
    "## compare Hurst exponent of closing prices with Brownian noise\n",
    "\n",
    "def hurst_exponent(time_series, max_lag=20):\n",
    "    \"\"\"Returns the Hurst Exponent of the time series\"\"\"\n",
    "    lags = range(2, max_lag)\n",
    "    # variances of the lagged differences\n",
    "    tau = [np.std(np.subtract(time_series[lag:], time_series[:-lag])) for lag in lags]\n",
    "    # calculate the slope of the log plot\n",
    "    reg = np.polyfit(np.log(lags), np.log(tau), 1)\n",
    "    return reg[0]\n",
    "\n",
    "for lag in [20, 100, 300, 500, 1000]:\n",
    "    hurst_exp = hurst_exponent(df['close'].to_numpy(), lag)\n",
    "    hurst_exp2 = hurst_exponent(brown_noise, lag)\n",
    "    print(f\"Hurst exponent with {lag} lags: {hurst_exp:.3f} \\t(brown noise = {hurst_exp2:.3f})\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The Hurst exponent relates to the fractal dimension of a series.  The entire closing price series has a similar Hurst exponent to brown noise/random walks, although differences may be seen when looking at shorter timespans.  A rolling window method (with resampling to give a more robust estimate) can discover this... "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## compare short-term Hurst exponents of closing prices with Brownian noise\n",
    " \n",
    "def hurst_fd(price_series, min_window=10, max_window=100, num_windows=20, num_samples=100):\n",
    "    \"\"\" function to calculate Hurst exponent. from https://archive.is/6lD77 \"\"\"\n",
    "    log_returns = np.diff(np.log(price_series))\n",
    "    window_sizes = np.linspace(min_window, max_window, num_windows, dtype=int)\n",
    "    R_S = []\n",
    "    \n",
    "    # for each window size, calculate R/S values\n",
    "    for w in window_sizes:\n",
    "        R, S = [], []\n",
    "        # resample multiple sequences for a more robust estimate\n",
    "        for _ in range(num_samples):\n",
    "            start = np.random.randint(0, len(log_returns) - w)\n",
    "            seq = log_returns[start:start + w]\n",
    "            R.append(np.max(seq) - np.min(seq))\n",
    "            S.append(np.std(seq))\n",
    "        R_S.append(np.mean(R) / np.mean(S))\n",
    "\n",
    "    # calculate the slope (Hurst Exponent) in a log-log plot\n",
    "    log_window_sizes = np.log(window_sizes)\n",
    "    log_R_S = np.log(R_S)\n",
    "    coeffs = np.polyfit(log_window_sizes, log_R_S, 1)\n",
    "\n",
    "    # extract Hurst Exponent \n",
    "    hurst_exponent = coeffs[0]\n",
    "\n",
    "    return hurst_exponent\n",
    "\n",
    "\n",
    "def rolling_hurst(price_series, window, min_window=10, max_window=100, num_windows=20, num_samples=25):\n",
    "    \"\"\" function to calculate rolling Hurst exponent \"\"\"\n",
    "    return price_series.rolling(window=window).apply(lambda x: hurst_fd(x, min_window, max_window, num_windows, num_samples), raw=True)\n",
    "\n",
    "\n",
    "df = pd.read_csv('data/OHLC_1h.csv', header=0, sep=',')\n",
    "ng = NoiseGenerator()\n",
    "brown_noise = ng.generate(dt=1e-5, n=10000, colour=ng.brown()) * 1000\n",
    "\n",
    "# generate graph\n",
    "ROLLING_WINDOW = 120\n",
    "df_hurst = pd.DataFrame(brown_noise[:3500])\n",
    "df_hurst['brown noise'] = rolling_hurst(df_hurst, ROLLING_WINDOW)\n",
    "df_hurst['close_price_1h'] = rolling_hurst(df['close'[:3500]], ROLLING_WINDOW)\n",
    "df_hurst = df_hurst.dropna(axis=0).reset_index(drop=True)\n",
    "df_hurst[:3000].plot(y=['brown noise', 'close_price_1h'], figsize=(8, 3))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When looking at shorter timespans the hourly closing prices show less tendency towards mean-reversion than a true random walk:  Prices have momentum, _i.e._ short-term trends are more prominent, in comparison.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Alternative Method To Get Price Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                              open     high      low    close    volume hour                 date        unix\n",
      "end_time_utc                                                                                                 \n",
      "2024-10-01 16:00:00+00:00  62558.9  62759.7  61777.0  62598.5  86170700   16  2024-10-01 16:00:00  1727798400\n",
      "2024-10-01 17:00:00+00:00  62598.5  62734.8  61849.0  62404.0  31476200   17  2024-10-01 17:00:00  1727802000\n",
      "2024-10-01 18:00:00+00:00  62404.0  62398.9  61015.0  62075.1  48412800   18  2024-10-01 18:00:00  1727805600\n",
      "2024-10-01 19:00:00+00:00  62075.1  62384.4  61794.0  61925.3  25741700   19  2024-10-01 19:00:00  1727809200\n",
      "2024-10-01 20:00:00+00:00  61925.3  62049.6  61536.7  61691.6  12668700   20  2024-10-01 20:00:00  1727812800\n",
      "2024-10-01 21:00:00+00:00  61691.6  61720.0  60121.0  60762.9  84900500   21  2024-10-01 21:00:00  1727816400\n",
      "2024-10-01 22:00:00+00:00  60762.9  61177.4  60602.8  61108.9  12091200   22  2024-10-01 22:00:00  1727820000\n",
      "2024-10-01 23:00:00+00:00  61108.9  61096.1  60634.3  60716.8   6150300   23  2024-10-01 23:00:00  1727823600\n",
      "                                        open     high      low    close hour     volume        unix                 date\n",
      "start_time_utc                                                                                                          \n",
      "2024-10-01 16:00:00.000000001+00:00  62598.5  62734.8  61015.0  61691.6   17  118299400  1727802000  2024-10-01 17:00:00\n",
      "2024-10-01 20:00:00.000000001+00:00  61691.6  61720.0  60121.0  60716.8   21  103142000  1727816400  2024-10-01 21:00:00\n",
      "                              open     high      low    close    volume hour                 date        unix\n",
      "end_time_utc                                                                                                 \n",
      "2024-10-01 16:00:00+00:00  62558.9  62759.7  61777.0  62598.5  86170700   16  2024-10-01 16:00:00  1727798400\n",
      "2024-10-01 17:00:00+00:00  62598.5  62734.8  61849.0  62404.0  31476200   17  2024-10-01 17:00:00  1727802000\n",
      "2024-10-01 18:00:00+00:00  62404.0  62398.9  61015.0  62075.1  48412800   18  2024-10-01 18:00:00  1727805600\n",
      "2024-10-01 19:00:00+00:00  62075.1  62384.4  61794.0  61925.3  25741700   19  2024-10-01 19:00:00  1727809200\n",
      "2024-10-01 20:00:00+00:00  61925.3  62049.6  61536.7  61691.6  12668700   20  2024-10-01 20:00:00  1727812800\n",
      "2024-10-01 21:00:00+00:00  61691.6  61720.0  60121.0  60762.9  84900500   21  2024-10-01 21:00:00  1727816400\n",
      "2024-10-01 22:00:00+00:00  60762.9  61177.4  60602.8  61108.9  12091200   22  2024-10-01 22:00:00  1727820000\n",
      "2024-10-01 23:00:00+00:00  61108.9  61096.1  60634.3  60716.8   6150300   23  2024-10-01 23:00:00  1727823600\n"
     ]
    }
   ],
   "source": [
    "import requests\n",
    "import time\n",
    "import datetime\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import calendar\n",
    "\n",
    "symbol = 'XBTUSD'\n",
    "buckets = '1h'           # 1m,5m,1h,1d\n",
    "allow_partials = False\n",
    "data_length = 10000\n",
    "resampled_buckets = '4h'  \n",
    "\n",
    "response = requests.get(f'https://www.bitmex.com/api/v1/trade/bucketed?binSize={buckets}&partial={allow_partials}&symbol={symbol}&count={data_length}&reverse=true').json()\n",
    "\n",
    "df = pd.DataFrame.from_dict(response)\n",
    "df = df.iloc[::-1]   # reverse dataframe\n",
    "\n",
    "df['hour'] = df.timestamp.str[11:13]\n",
    "df['end_time_utc'] = pd.to_datetime(df['timestamp'], utc=False)\n",
    "df['date'] = df.timestamp.str[0:10] + ' ' + df.timestamp.str[11:19]\n",
    "df['unix'] = df.date.apply(lambda string : calendar.timegm(time.strptime(string, \"%Y-%m-%d %H:%M:%S\")))\n",
    "df = df.set_index('end_time_utc')\n",
    "df = df.drop(columns=['vwap', 'lastSize', 'turnover', \n",
    "                    'homeNotional', 'foreignNotional', \n",
    "                    'trades', 'symbol', 'timestamp'], axis=1)\n",
    "\n",
    "if resampled_buckets:\n",
    "\n",
    "    if allow_partials == 'false':\n",
    "        # remove recent buckets that will form a partial bucket if resampled  \n",
    "        while True:\n",
    "            if int(df.tail(1)['hour']) % 12 == 0:\n",
    "                break\n",
    "            else: \n",
    "                df.drop(df.tail(1).index, inplace=True)\n",
    "\n",
    "    ohlc_dict = {'open':'first', 'high':'max', 'low':'min', 'close':'last', 'hour':'first', 'volume':'sum', 'unix':'first', 'date':'first'}\n",
    "    df2 = df.resample(resampled_buckets, offset=1).apply(ohlc_dict).dropna(how='any')\n",
    "    df2.index.names = ['start_time_utc']\n",
    "    # df2 = df2.drop(columns=['hour'], axis=1) \n",
    "\n",
    "    print(df.tail(8))\n",
    "    print(df2.tail(2))\n",
    "\n",
    "else:\n",
    "    print(df.tail(8))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "## Feature Engineering\n",
    "#### 1. Create categorical labels."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "         unix                 date     open     high      low    close       volume  volume USD        ATR target  day  hour  year\n",
      "0  1526364000  2018-05-15 06:00:00  8733.86  8796.68  8707.28  8740.99   4906603.14      559.93  89.400000   down    4     6  2018\n",
      "1  1526367600  2018-05-15 07:00:00  8740.99  8766.00  8721.11  8739.00   2390398.89      273.58  86.220714   flat    4     7  2018\n",
      "2  1526371200  2018-05-15 08:00:00  8739.00  8750.27  8660.53  8728.49   7986062.84      917.79  86.472092     up    4     8  2018\n",
      "3  1526374800  2018-05-15 09:00:00  8728.49  8754.40  8701.35  8708.32   1593991.98      182.62  84.084800   flat    4     9  2018\n",
      "4  1526378400  2018-05-15 10:00:00  8708.32  8865.00  8695.11  8795.90  11101273.74     1260.69  90.213742   down    4    10  2018\n",
      "COUNT: 50515\n",
      " down     6850\n",
      "flat    36154\n",
      "up       7509\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "import numpy as np  \n",
    "import pandas as pd\n",
    "\n",
    "\n",
    "def convert_unix_time_into_day(seconds):\n",
    "    \"\"\" returns day of week as category label \"\"\"\n",
    "    day_of_week = time.strftime('%A', time.localtime(seconds))\n",
    "    dict = {'Saturday':1, 'Sunday':2, 'Monday':3, 'Tuesday':4, 'Wednesday':5, 'Thursday':6, 'Friday':7}\n",
    "    day_code = dict[day_of_week]\n",
    "    return day_code\n",
    "\n",
    "\n",
    "def wwma(pd_series, period):\n",
    "    \"\"\" w. wilder's exponential moving average \"\"\"\n",
    "    return pd_series.ewm(alpha=1/period, adjust=False, ignore_na=True).mean()\n",
    "\n",
    "\n",
    "def atr(df, length=14):\n",
    "    \"\"\" average true range (for column with latest values at top) \"\"\"\n",
    "    df_high, df_low, df_prev_close = df['high'], df['low'], df['close'].shift()\n",
    "    df_tr = [df_high - df_low, df_high - df_prev_close, df_low - df_prev_close]\n",
    "    df_tr = [tr.abs() for tr in df_tr]\n",
    "    df_tr = pd.concat(df_tr, axis=1).max(axis=1)\n",
    "    df_atr = wwma(df_tr, length)\n",
    "    return df_atr\n",
    "\n",
    "\n",
    "PERCENT_CHANGE = False    # Must be false - adding NaNs?\n",
    "PERIOD = 30                 \n",
    "PERIOD_FAST = 11          \n",
    "ISHIMOKU_SCALE = 4\n",
    "WINDOW = 20\n",
    "STEP = 2\n",
    "COMMISSION_FRAC = 0.001\n",
    "COMMISSION_MULT = 5  \n",
    "ATR_MULT = 0.25\n",
    "RUN_ID = '6'\n",
    "\n",
    "# load price data into dataframe and reorder to most recent at bottom\n",
    "df = pd.read_csv('data/OHLC_1h.csv', header=0, sep=',')\n",
    "df = df.sort_index(ascending=False, ignore_index=True)\n",
    "\n",
    "# calculate average true range\n",
    "df_atr = atr(df)\n",
    "df['ATR'] = df_atr\n",
    "\n",
    "# create target labels \n",
    "y_labels =[]\n",
    "\n",
    "for i in range(2,len(df)):\n",
    "\n",
    "    # up condition: (high or high[1] > close[2] + spread) and (low and low[1] > low[2] + spread)     \n",
    "    if (\n",
    "        ( df.loc[i, 'high'] > (df.loc[i-2, 'close'] + ATR_MULT * df.loc[i, 'ATR'])  or  \n",
    "        df.loc[i-1, 'high'] > (df.loc[i-2, 'close'] + ATR_MULT * df.loc[i, 'ATR']) )  and \n",
    "\n",
    "        ( df.loc[i, 'high'] > (df.loc[i-2, 'close'] * (1 + COMMISSION_FRAC * COMMISSION_MULT))  or  \n",
    "        df.loc[i-1, 'high'] > (df.loc[i-2, 'close'] * (1 + COMMISSION_FRAC * COMMISSION_MULT)) )  and\n",
    "\n",
    "        ( df.loc[i, 'low'] > (df.loc[i-2, 'low'] + ATR_MULT * df.loc[i, 'ATR'])  and  \n",
    "        df.loc[i-1, 'low'] > (df.loc[i-2, 'low'] + ATR_MULT * df.loc[i, 'ATR']) )\n",
    "    ):\n",
    "        label = 'up'\n",
    "\n",
    "    # down condition: (low or low[1] < close[2] - spread) and (high and high[1] < high[2] - spread)    \n",
    "    elif (\n",
    "        ( df.loc[i, 'low'] < (df.loc[i-2, 'close'] - ATR_MULT * df.loc[i, 'ATR'])  or  \n",
    "          df.loc[i-1, 'low'] < (df.loc[i-2, 'close'] - ATR_MULT * df.loc[i, 'ATR']) )  and \n",
    "          \n",
    "        ( df.loc[i, 'low'] < (df.loc[i-2, 'close'] * (1 - COMMISSION_FRAC * COMMISSION_MULT))  or  \n",
    "          df.loc[i-1, 'low'] < (df.loc[i-2, 'close'] * (1 - COMMISSION_FRAC * COMMISSION_MULT)) )  and\n",
    "\n",
    "        ( df.loc[i, 'high'] < (df.loc[i-2, 'high'] - ATR_MULT * df.loc[i, 'ATR'])  and  \n",
    "          df.loc[i-1, 'high'] < (df.loc[i-2, 'high'] - ATR_MULT * df.loc[i, 'ATR']) )\n",
    "    ):\n",
    "        label = 'down'\n",
    "    \n",
    "    else:\n",
    "        label = 'flat'\n",
    "\n",
    "    y_labels.append(label)\n",
    "\n",
    "y_labels.extend([float('NaN'), float('NaN')])\n",
    "\n",
    "# add labels to new dataframe column\n",
    "df['target'] = y_labels\n",
    "\n",
    "# create day and hour categories\n",
    "df['day'] = df['unix'].map(lambda x: convert_unix_time_into_day(x))\n",
    "df['hour'] = df['date'].str.slice(start=11, stop=13).apply(pd.to_numeric)\n",
    "df['year'] = df['date'].str.slice(start=0, stop=4).apply(pd.to_numeric)\n",
    "\n",
    "pd.set_option(\"expand_frame_repr\", False)\n",
    "pd.set_option(\"display.max_columns\", 100)\n",
    "print(df.head())\n",
    "\n",
    "df_ylabels = df['target'].copy()\n",
    "df_ylabels = pd.get_dummies(df_ylabels, columns=['target'])\n",
    "print(f'COUNT: {len(df)}\\n', df_ylabels.sum(axis=0))\n",
    "del df_ylabels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2. Create features using variance, filters, significant levels, volume indicators, and fractal dimension"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "         unix                 date     open     high     low    close      volume  volume USD         ATR target  day  hour  year       SMA_30       SMA_11  vol SMA_30  vol SMA_11     open_     high_      low_    close_      vol_   close_f     vol_f     HH22     LL22      chh_      cll_        bk     d_sdev    c-c1    h-h1   l-l1        v-v1    hlc-hlc1      vwp-vwp1  rr_wvap_fast      rr_wvap          obv       acc_dis           mfv    tenkan     kijun  senkou_a  senkou_b    senkou    d_frama  rel_frama  frac_dim  d_frac_dim      ngd_3      ngd_4      ngd_6      ngd_8      ngd_11      ngd_16       ngd_22       ngd_32     drsi_7    drsi_14    drsi_21\n",
      "0  1526468400  2018-05-16 11:00:00  8371.13  8439.44  8370.0  8373.00  2680259.87      319.29  104.106814   down    5    11  2018  8477.674667  8249.193636  592.550000  662.216364 -0.012568 -0.004510 -0.012701 -0.012347 -0.461159  0.015008 -0.517846  8771.02  8100.01 -0.045379 -0.033702  1.536641 -14.265967    1.87  -16.22  132.9 -4987234.43  117.303333 -9.685257e+10   8370.694027  8370.636398 -22020735.69 -7.651111e+06 -2.413908e+06  0.994605  0.991600  0.000119  1.009113 -1.008995   1.870000   1.000000  0.676550   -0.309439   1.356111  28.734731  52.347200 -79.173469 -296.162754 -147.514770   163.464996  1563.706205  28.107646   8.320180   8.272585\n",
      "1  1526472000  2018-05-16 12:00:00  8373.00  8399.10  8356.4  8365.08  3031532.02      362.01   99.720613   flat    5    12  2018  8465.144333  8252.720909  585.952667  597.769091 -0.010885 -0.007802 -0.012846 -0.011821 -0.382186  0.013615 -0.394398  8761.70  8100.01 -0.045267 -0.032725  1.380993   1.950352   -7.92  -40.34  -13.6   351272.15  -56.580000  6.713565e+09   8362.764955  8362.717235 -25052267.71 -9.408983e+06 -1.757872e+06  0.995994  0.990923  0.000119  1.008420 -1.008301  -3.946520   1.000475  1.151425    0.474875 -21.875830 -20.163856  43.223943 -43.771380 -215.778393  223.645640  -287.015957  1949.646241  26.693610   8.231494   8.194979\n",
      "2  1526475600  2018-05-16 13:00:00  8365.08  8381.73  8145.1  8179.99  8614269.80     1046.38  109.499855     up    5    13  2018  8446.510667  8256.953636  611.712667  537.421818 -0.009641 -0.007670 -0.035685 -0.031554  0.710574 -0.009321  0.947037  8589.10  8100.01 -0.047631 -0.009874 -1.076812  -2.659984 -185.09  -17.37 -211.3  5582737.78 -290.366667  1.066069e+11   8177.623066  8177.578446 -33666537.51 -1.545742e+07 -6.048441e+06  1.014718  1.011962  0.000124  1.030668 -1.030544 -18.906348   1.020802  1.578398    0.426972 -44.844179 -61.974341 -32.181569 -22.305413 -110.351664  495.000850  -677.209838  2227.017657  20.882666   7.789607   7.726372\n",
      "3  1526479200  2018-05-16 14:00:00  8179.99  8229.91  8173.6  8225.00  2936522.28      357.91  105.700579   flat    5    14  2018  8429.727667  8262.427273  593.050000  525.159091 -0.029626 -0.023704 -0.030384 -0.024286 -0.396493 -0.004530 -0.318473  8589.10  8100.01 -0.042391 -0.015431 -0.561038  -4.172543   45.01 -151.82   28.5 -5677747.52 -108.316667 -1.096345e+11   8222.647081  8222.606875 -30730015.23 -1.307531e+07  2.382113e+06  1.008179  1.007795  0.000123  1.014536 -1.014413 -12.514713   1.013694  1.543185   -0.035213  -3.727744 -24.078934 -79.165749 -14.582752   -9.616627  646.678859  -980.874143  2395.774031  34.970302  10.406715   9.950343\n",
      "4  1526482800  2018-05-16 15:00:00  8225.00  8377.29  8211.8  8285.80  4663319.71      561.20  109.971252   flat    5    15  2018  8415.643667  8267.556364  605.669333  518.839091 -0.022653 -0.004557 -0.024222 -0.015429 -0.073422  0.002207  0.081646  8583.60  8100.01 -0.034694 -0.022937  0.196629  -1.812801   60.80  147.38   38.2  1726797.43  205.846667  3.401977e+10   8283.462135  8283.426508 -26066695.52 -1.356520e+07 -4.898881e+05  0.998347  1.001760  0.000121  1.007091 -1.006970  -5.183242   1.005630  1.543185    0.000000  52.390076  57.526015 -25.841362  12.869726   84.011138  692.676093 -1170.799129  2467.183458  40.492242  12.498963  12.079363\n",
      "             unix                 date     open     high      low    close     volume    volume USD         ATR target  day  hour  year        SMA_30        SMA_11    vol SMA_30    vol SMA_11     open_     high_      low_    close_      vol_   close_f     vol_f     HH22     LL22      chh_      cll_        bk     d_sdev   c-c1   h-h1   l-l1       v-v1    hlc-hlc1      vwp-vwp1  rr_wvap_fast       rr_wvap           obv       acc_dis        mfv    tenkan     kijun  senkou_a  senkou_b    senkou     d_frama  rel_frama  frac_dim  d_frac_dim      ngd_3      ngd_4       ngd_6       ngd_8      ngd_11      ngd_16      ngd_22      ngd_32     drsi_7    drsi_14    drsi_21\n",
      "50479  1708192800  2024-02-17 18:00:00  51231.0  51299.0  51153.0  51278.0  30.546773  1.566377e+06  285.154080     up    1    18  2024  51778.600000  51380.454545  2.812646e+06  2.902317e+06 -0.010576 -0.009263 -0.012082 -0.009668 -0.443095 -0.001994 -0.460301  52251.0  50647.0 -0.018622 -0.012459 -0.289341 -34.642582   43.0   29.0  182.0  -8.407337  225.333333 -9.962746e+05  51275.673636  51275.644390 -9.373490e+08  2.295319e+09  21.611322  0.994052  1.000712  0.000019  1.003335 -1.003315   43.000000   1.000000  0.783519   -0.209802  73.774039   9.601855  190.201412  176.536097   33.980534  -77.211063 -228.814256 -227.242807  59.052878  55.816748  55.008611\n",
      "50480  1708196400  2024-02-17 19:00:00  51283.0  51448.0  51250.0  51414.0  32.900628  1.691553e+06  278.928788   flat    1    19  2024  51741.900000  51349.636364  2.676756e+06  2.908970e+06 -0.008869 -0.005680 -0.009507 -0.006337 -0.368059  0.001253 -0.418504  52251.0  50647.0 -0.016019 -0.015144  0.222915 -29.066402  136.0  149.0   97.0   2.353855  291.333333  2.909759e+05  51411.680073  51411.652955 -9.373489e+08  2.295319e+09  21.492873  0.993932  0.997530  0.000019  1.000681 -1.000661  136.000000   1.000000  0.978742    0.195224  -2.824810  17.173137   72.073470  213.355924  152.681909   19.063610 -164.922161 -205.981257  61.024747  57.735687  57.678819\n",
      "50481  1708200000  2024-02-17 20:00:00  51409.0  51581.0  51324.0  51528.0  22.796003  1.174632e+06  277.362446     up    1    20  2024  51727.633333  51333.181818  2.295022e+06  2.728981e+06 -0.006160 -0.002835 -0.007803 -0.003859 -0.488182  0.003795 -0.569571  52251.0  50647.0 -0.013837 -0.017395  0.691572 -22.559587  114.0  133.0   74.0 -10.104626  245.000000 -1.205313e+06  51525.687858  51525.662010 -9.373489e+08  2.295319e+09  13.341847  0.995109  0.994226  0.000019  0.998467 -0.998448   99.146369   0.999712  1.030348    0.051606  50.652286  56.416919  -20.254905  166.396909  229.194090  116.839349  -81.781217 -169.259622  60.728065  58.475063  59.137367\n",
      "50482  1708203600  2024-02-17 21:00:00  51549.0  52019.0  51475.0  51811.0  88.026180  4.560724e+06  296.407986   flat    1    21  2024  51724.433333  51340.181818  2.334198e+06  2.753947e+06 -0.003392  0.005695 -0.004822  0.001674  0.953872  0.009171  0.656068  52251.0  50647.0 -0.008421 -0.022983  1.524859  13.296872  283.0  438.0  151.0  65.230177  683.333333  7.893056e+06  51808.691855  51808.674776 -9.373488e+08  2.295319e+09  20.674039  0.995657  0.990774  0.000019  0.993013 -0.992994  266.316987   0.999391  1.024329   -0.006019  88.339412  88.609089    6.083141  102.414885  265.540992  212.822774   20.106115 -114.047375  67.728343  62.454425  63.485118\n",
      "50483  1708207200  2024-02-17 22:00:00  51806.0  51870.0  51598.0  51725.0  57.006645  2.948669e+06  294.664558   flat    1    22  2024  51713.166667  51341.545455  2.226948e+06  2.947949e+06  0.001795  0.003033 -0.002227  0.000229  0.324085  0.007469  0.000244  52210.0  50647.0 -0.009289 -0.021285  1.259795  -0.987303  -86.0 -149.0  123.0 -31.019535  -54.666667 -3.749170e+06  51722.681728  51722.672501 -9.373489e+08  2.295319e+09  -3.758680  0.998250  0.992421  0.000019  0.994664 -0.994645  -31.359873   1.000447  1.120000    0.095670 -19.878957  19.566013   52.703819   31.074380  239.053058  279.137043  120.086241  -51.345084  48.885358  55.010372  57.293778\n",
      "50484\n"
     ]
    }
   ],
   "source": [
    "\n",
    "import math\n",
    "#import numba\n",
    "\n",
    "\n",
    "def bollinger_k(df, length=20):\n",
    "    \"\"\" bollinger coeffient and delta standard deviation\"\"\"\n",
    "    df_tp = (df['high'] + df['low'] + df['close']) / 3\n",
    "    df_sma = df_tp.rolling(length).mean()\n",
    "    df_sdev = df_tp.rolling(length).std()\n",
    "    df_delta_sdev = df_sdev.diff()\n",
    "    df_bk = (df['close'] - df_sma) / df_sdev \n",
    "    return df_bk, df_delta_sdev\n",
    "\n",
    "\n",
    "def rr_vwap(dataframe, length):\n",
    "    \"\"\" relative rolling volume-weighted average price \"\"\"\n",
    "    df = dataframe.copy()\n",
    "    df['period_total'] = (df['high'] + df['low'] + df['close'] / 3) * df['volume']\n",
    "    df['cum_tot'] = df['period_total'].rolling(length).sum()\n",
    "    df['cum_vol'] = df['volume'].rolling(length).sum()\n",
    "    df_rvwap = df['close'] - (df['cum_tot'] / df['cum_vol']) / df['close']\n",
    "    return df_rvwap\n",
    "\n",
    "\n",
    "# @numba.jit\n",
    "def frama(dataframe, batch=10):\n",
    "    \"\"\" fractal adaptive moving average \"\"\"\n",
    "    df = dataframe.copy() \n",
    "    price = df.close  \n",
    "    fractal_dims = []\n",
    "    filtered_prices = np.array(price)\n",
    "\n",
    "    for i in range(0, len(df)):  \n",
    "        \n",
    "        if i < 2 * batch:  \n",
    "            fractal_dims.append(np.nan)\n",
    "            continue  \n",
    "\n",
    "        v1 = price[i-2*batch : i - batch]  \n",
    "        v2 = price[i - batch : i]\n",
    "        H1 = np.max(v1)  \n",
    "        L1 = np.min(v1)  \n",
    "        N1 = (H1 - L1) / batch\n",
    "        H2 = np.max(v2)  \n",
    "        L2 = np.min(v2)  \n",
    "        N2 = (H2 - L2) / batch\n",
    "        H = np.max([H1, H2])  \n",
    "        L = np.min([L1, L2])  \n",
    "        N3 = (H - L) / (2 * batch)\n",
    "\n",
    "        fractal_dim = 0  \n",
    "        if N1 > 0 and N2 > 0 and N3 > 0:  \n",
    "            fractal_dim = (np.log(N1 + N2) - np.log(N3)) / np.log(2)\n",
    "        fractal_dims.append(fractal_dim)\n",
    "\n",
    "        alpha = np.exp(-4.6 * (fractal_dim - 1))    # 4.6 limits slow EMA length < 200 \n",
    "        alpha = np.max([alpha, 0.1])  \n",
    "        alpha = np.min([alpha, 1])\n",
    "\n",
    "        filtered_prices[i] = alpha * price[i] + (1 - alpha) * filtered_prices[i-1]    # single-pole lowpass filter\n",
    "\n",
    "    df['frama'] = filtered_prices\n",
    "    df['d_frama'] = df['frama'] - df['frama'].shift(1)\n",
    "    df['rel_frama'] = df['frama'] / df['close']\n",
    "    df['frac_dim'] = fractal_dims\n",
    "    df['d_frac_dim'] = df['frac_dim'] - df['frac_dim'].shift(1)\n",
    "\n",
    "    return df.d_frama, df.rel_frama, df.frac_dim, df.d_frac_dim\n",
    "\n",
    "\n",
    "# @numba.jit\n",
    "def z_transformer(dataframe, mode, period):\n",
    "    \"\"\" digital signal processor based on a generalised z-domain transfer function \"\"\"\n",
    "    df = dataframe.copy() \n",
    "    price = df.close  \n",
    "    transformed_prices = np.array(price)\n",
    "\n",
    "    pi = 3.14159\n",
    "    c1, N, b1, b2, a1, a2 = 0.0, 0.0, 0.0, 0.0, 0.0, 0.0\n",
    "    b0 = 1.0\n",
    "\n",
    "    if mode == 'highpass':\n",
    "        threshold_period = period\n",
    "        alpha = (math.cos(2.0*pi/threshold_period) + math.sin(2.0*pi/threshold_period) - 1) / math.cos(2.0*pi/threshold_period)\n",
    "        c0 = 1.0 - alpha / 2.0\n",
    "        b1 = -1.0\n",
    "        a1 = 1.0 - alpha\n",
    "\n",
    "    if mode == 'bandpass':\n",
    "        centre_period = period\n",
    "        delta = 0.2    # half-bandwidth fraction: min=0.05, max=0.5\n",
    "        beta = math.cos(2.0*pi / centre_period)\n",
    "        gamma = 1.0 / math.cos(4.0*pi*delta / centre_period)\n",
    "        alpha = gamma - math.sqrt(gamma**2 - 1.0)\n",
    "        c0 = (1.0 - alpha) / 2.0\n",
    "        b2 = -1.0\n",
    "        a1 = beta * (alpha + 1.0)\n",
    "        a2 = -alpha\n",
    "\n",
    "    for i in range(0, len(df)):  \n",
    "        if i < 2:  \n",
    "            continue  \n",
    "\n",
    "        transformed_prices[i] = c0 * (b0*price[i] + b1*price[i-1] + b2*price[i-2]) + a1*transformed_prices[i-1] + a2*transformed_prices[i-2] # - c1*price[N]\n",
    "\n",
    "    df['output'] = transformed_prices\n",
    "    df['delta_output'] = df.output - df.output.shift(1)\n",
    "    df['relative_output'] = df.output / df.close\n",
    "\n",
    "    return df.output, df.delta_output, df.relative_output\n",
    "\n",
    "\n",
    "def ema(pd_series, period):\n",
    "    \"\"\" exponential moving average \"\"\"\n",
    "    return pd_series.ewm(alpha=1/period, adjust=False, ignore_na=True).mean()\n",
    "\n",
    "\n",
    "# @numba.jit\n",
    "def rma(x, n):\n",
    "    \"\"\" running moving average \"\"\"\n",
    "    a = np.full_like(x, np.nan)\n",
    "    a[n] = x[1:n+1].mean()\n",
    "    for i in range(n+1, len(x)):\n",
    "        a[i] = (a[i-1] * (n - 1) + x[i]) / n\n",
    "    return a\n",
    "\n",
    "\n",
    "def detrended_rsi(dataframe, hp_threshold=8, rsi_period=14):\n",
    "    \"\"\" relative strength index of high-frequency component of price \"\"\"\n",
    "    df = dataframe.copy()\n",
    "\n",
    "    df['detrended_price'] = z_transformer(df, mode='highpass', period=hp_threshold)[0]\n",
    "    df['change'] = df['detrended_price'].diff()\n",
    "    df['gain'] = df.change.mask(df.change < 0, 0.0)\n",
    "    df['loss'] = -df.change.mask(df.change > 0, -0.0)\n",
    "\n",
    "    df['avg_gain'] = rma(df.gain.to_numpy(), rsi_period)\n",
    "    df['avg_loss'] = rma(df.loss.to_numpy(), rsi_period)\n",
    "\n",
    "    df['rs'] = df.avg_gain / df.avg_loss\n",
    "    df['rsi'] = 100 - (100 / (1 + df.rs))\n",
    "\n",
    "    return df.rsi\n",
    "\n",
    "\n",
    "def acc_dis_mfv(dataframe):\n",
    "    \"\"\" accumulation-distribution indicator \"\"\"\n",
    "    df = dataframe.copy()\n",
    "    price = df.close  \n",
    "    acc_dis = np.array(price)\n",
    "    df['mfv'] = df.volume * ((df.close - df.low) - (df.high - df.close)) / (df.high - df.low + 1)\n",
    "    mfv = df['mfv'].to_numpy()\n",
    "\n",
    "    for i in range(0, len(df)):  \n",
    "        if i < 1:  \n",
    "            continue  \n",
    "\n",
    "        # if acc_dis[i-1] == np.nan:\n",
    "        #     acc_dis[i-1] = price[i-1]\n",
    "\n",
    "        acc_dis[i] = acc_dis[i-1] + mfv[i]\n",
    "    df['acc_dis'] = acc_dis\n",
    "\n",
    "    return df.acc_dis, df.mfv\n",
    "\n",
    "\n",
    "def obv(dataframe):\n",
    "    \"\"\" on-balance-volume indicator \"\"\"\n",
    "    df = dataframe.copy()\n",
    "    volume = df.volume\n",
    "    price = df.close  \n",
    "    obv = np.array(volume)\n",
    "\n",
    "    for i in range(0, len(df)):  \n",
    "        if i < 1:  \n",
    "            continue  \n",
    "        if price[i] > price[i-1]: new = volume[i]\n",
    "        if price[i] == price[i-1]: new = 0.0\n",
    "        if price[i] < price[i-1]: new = -volume[i]\n",
    "        obv[i] = obv[i-1] + new\n",
    "\n",
    "    df['obv'] = obv\n",
    "    return df.obv\n",
    "\n",
    "\n",
    "# calculate simple moving averages of closing price\n",
    "df[f'SMA_{PERIOD}'] = df['close'].rolling(PERIOD).mean()\n",
    "df[f'SMA_{PERIOD_FAST}'] = df['close'].rolling(PERIOD_FAST).mean()\n",
    "\n",
    "# calculate simple moving averages of volume\n",
    "df[f'vol SMA_{PERIOD}'] = df['volume USD'].rolling(PERIOD).mean()\n",
    "df[f'vol SMA_{PERIOD_FAST}'] = df['volume USD'].rolling(PERIOD_FAST).mean()\n",
    "\n",
    "# centralise price and volume data around relevant slower SMA\n",
    "df['open_'] = (df['open'] - df[f'SMA_{PERIOD}']) / df[f'SMA_{PERIOD}']\n",
    "df['high_'] = (df['high'] - df[f'SMA_{PERIOD}']) / df[f'SMA_{PERIOD}']\n",
    "df['low_'] = (df['low'] - df[f'SMA_{PERIOD}']) / df[f'SMA_{PERIOD}']\n",
    "df['close_'] = (df['close'] - df[f'SMA_{PERIOD}']) / df[f'SMA_{PERIOD}']\n",
    "df['vol_'] = (df['volume USD'] - df[f'vol SMA_{PERIOD}']) / df[f'vol SMA_{PERIOD}']\n",
    "\n",
    "# centralise price and volume data around relevant faster SMA\n",
    "df['close_f'] = (df['close'] - df[f'SMA_{PERIOD_FAST}']) / df[f'SMA_{PERIOD_FAST}']\n",
    "df['vol_f'] = (df['volume USD'] - df[f'vol SMA_{PERIOD_FAST}']) / df[f'vol SMA_{PERIOD_FAST}']\n",
    "\n",
    "# calculate highest high and lowest low in the last 'PERIOD_FAST*2' prices\n",
    "df[f'HH{PERIOD_FAST*2}'] = df['high'].rolling(PERIOD_FAST*2).max().shift()\n",
    "df[f'LL{PERIOD_FAST*2}'] = df['low'].rolling(PERIOD_FAST*2).min().shift()\n",
    "\n",
    "# close price fraction of highest high and lowest low \n",
    "df['chh_'] = (df['close'] - df[f'HH{PERIOD_FAST*2}']) / df[f'HH{PERIOD_FAST*2}']\n",
    "df['cll_'] = (df[f'LL{PERIOD_FAST*2}'] - df['close']) / df[f'LL{PERIOD_FAST*2}']\n",
    "\n",
    "# calculate bollinger coeffient and standard deviation add as column to dataframe\n",
    "df['bk'], df['d_sdev'] = bollinger_k(df, length=PERIOD_FAST)\n",
    "\n",
    "# deltas\n",
    "df['c-c1'] = df['close'] - df['close'].shift(1)\n",
    "df['h-h1'] = df['high'] - df['high'].shift(1)\n",
    "df['l-l1'] = df['low'] - df['low'].shift(1)\n",
    "df['v-v1'] = df['volume'] - df['volume'].shift(1)\n",
    "df['hlc-hlc1'] = (df['high'] + df['low'] + df['close'] / 3) - (df['high'].shift(1) + df['low'].shift(1) + df['close'].shift(1) / 3)\n",
    "df['vwp-vwp1'] = ((df['high'] + df['low'] + df['close'] / 3) * df['volume']) - ((df['high'].shift(1) + df['low'].shift(1) + df['close'].shift(1) / 3) * df['volume'].shift(1))\n",
    "\n",
    "# relative volume-weighted average price\n",
    "df['rr_wvap_fast'] = rr_vwap(df, length=7)\n",
    "df['rr_wvap'] = rr_vwap(df, length=PERIOD)\n",
    "\n",
    "# volume indicators\n",
    "df['obv'] = obv(df)\n",
    "df['acc_dis'], df['mfv'] = acc_dis_mfv(df)\n",
    "\n",
    "# ishimoku cloud relative to close price\n",
    "df['tenkan'] = ((df['high'].rolling(ISHIMOKU_SCALE * 1).max() + df['low'].rolling(ISHIMOKU_SCALE * 1).min()) / 2) / df['close']\n",
    "df['kijun'] = ((df['high'].rolling(ISHIMOKU_SCALE * 3).max() + df['low'].rolling(ISHIMOKU_SCALE * 3).min()) / 2) / df['close']\n",
    "df['senkou_a'] = (df['tenkan'] + df['kijun']) / 2\n",
    "df['senkou_b'] = (df['high'].rolling(ISHIMOKU_SCALE * 6).max() + df['low'].rolling(ISHIMOKU_SCALE * 6).min()) / 2\n",
    "df['senkou'] = (df['senkou_a'] - df['senkou_b']) / df['close']\n",
    "df['senkou_a'] = df['senkou_a'] / df['close']\n",
    "df['senkou_b'] = df['senkou_b'] / df['close']\n",
    "\n",
    "# fractal adaptive moving average\n",
    "df['d_frama'], df['rel_frama'], df['frac_dim'], df['d_frac_dim'] = frama(df, batch=10)\n",
    "\n",
    "# negative group delay bandpass array\n",
    "df['ngd_3'] = z_transformer(df, mode='bandpass', period=3)[0]   \n",
    "df['ngd_4'] = z_transformer(df, mode='bandpass', period=4)[0]   \n",
    "df['ngd_6'] = z_transformer(df, mode='bandpass', period=6)[0]      \n",
    "df['ngd_8'] = z_transformer(df, mode='bandpass', period=8)[0]     \n",
    "df['ngd_11'] = z_transformer(df, mode='bandpass', period=11)[0]    \n",
    "df['ngd_16'] = z_transformer(df, mode='bandpass', period=16)[0]    \n",
    "df['ngd_22'] = z_transformer(df, mode='bandpass', period=22)[0]   \n",
    "df['ngd_32'] = z_transformer(df, mode='bandpass', period=32)[0]   \n",
    "\n",
    "# detrended relative strength index\n",
    "df['drsi_7'] = detrended_rsi(df, hp_threshold=7, rsi_period=7)\n",
    "df['drsi_14'] = detrended_rsi(df, hp_threshold=14, rsi_period=14)\n",
    "df['drsi_21'] = detrended_rsi(df, hp_threshold=21, rsi_period=14)\n",
    "\n",
    "# # capture indices of NaN rows to also remove from outcomes list\n",
    "# nan_indices = df.index[df.isnull().any(axis=1)].tolist()\n",
    "\n",
    "df = df.dropna(axis=0)\n",
    "df.reset_index(drop=True, inplace=True)\n",
    "\n",
    "# check dataframe\n",
    "print(df.head())\n",
    "print(df.tail())\n",
    "print(len(df))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3. Calculate profit/loss fraction for each trade direction and timestamp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "25232\n",
      "25232\n",
      "25232\n",
      "25232\n"
     ]
    }
   ],
   "source": [
    "import random\n",
    "\n",
    "WINDOW = 20\n",
    "STEP = 2\n",
    "ATR_MULT_TP = 0.25\n",
    "ATR_MULT_SL = 0.25\n",
    "COMMISSION_FRAC = 0.001\n",
    "PL_ID = 'b'\n",
    "\n",
    "up_outcomes = []\n",
    "up_outcomes_passive = []\n",
    "\n",
    "# tp_frac = ((df.loc[i, 'close'] + (ATR_MULT * df.loc[i, 'ATR'])) / df.loc[i, 'close']) - COMMISSION_FRAC\n",
    "# sl_frac = ((df.loc[i, 'low'] - ATR_MULT * df.loc[i, 'ATR']) / df.loc[i, 'close']) - COMMISSION_FRAC\n",
    "\n",
    "for i in range(len(df)-2):\n",
    "\n",
    "    if bool(random.getrandbits(1)) == True:    # randomises order of whether tp or sl is hit first if both are crossed\n",
    "\n",
    "        if df.loc[i+1, 'low'] < (df.loc[i, 'close'] - ATR_MULT_SL * df.loc[i, 'ATR']):\n",
    "            up_outcomes.append(((df.loc[i, 'close'] - ATR_MULT_SL * df.loc[i, 'ATR']) / df.loc[i, 'close']) - COMMISSION_FRAC)  \n",
    "\n",
    "        elif df.loc[i+1, 'high'] > (df.loc[i, 'close'] + ATR_MULT_TP * df.loc[i, 'ATR']):    \n",
    "            up_outcomes.append(((df.loc[i, 'close'] + (ATR_MULT_TP * df.loc[i, 'ATR'])) / df.loc[i, 'close']) - COMMISSION_FRAC)\n",
    "\n",
    "        elif df.loc[i+2, 'low'] < (df.loc[i, 'close'] - ATR_MULT_SL * df.loc[i, 'ATR']):\n",
    "            up_outcomes.append(((df.loc[i, 'close'] - ATR_MULT_SL * df.loc[i, 'ATR']) / df.loc[i, 'close']) - COMMISSION_FRAC)\n",
    "\n",
    "        elif df.loc[i+2, 'high'] > (df.loc[i, 'close'] + ATR_MULT_TP * df.loc[i, 'ATR']):\n",
    "            up_outcomes.append(((df.loc[i, 'close'] + (ATR_MULT_TP * df.loc[i, 'ATR'])) / df.loc[i, 'close']) - COMMISSION_FRAC)\n",
    "\n",
    "        else:\n",
    "            up_outcomes.append((df.loc[i+2, 'close'] / df.loc[i, 'close']) - COMMISSION_FRAC)\n",
    "\n",
    "        up_outcomes_passive.append((df.loc[i+2, 'close'] / df.loc[i, 'close']) - (1.5 * COMMISSION_FRAC))\n",
    "\n",
    "    else:\n",
    "        if df.loc[i+1, 'high'] > (df.loc[i, 'close'] + ATR_MULT_TP * df.loc[i, 'ATR']):    \n",
    "            up_outcomes.append(((df.loc[i, 'close'] + (ATR_MULT_TP * df.loc[i, 'ATR'])) / df.loc[i, 'close']) - COMMISSION_FRAC)      \n",
    "\n",
    "        elif df.loc[i+1, 'low'] < (df.loc[i, 'close'] - ATR_MULT_SL * df.loc[i, 'ATR']):\n",
    "            up_outcomes.append(((df.loc[i, 'close'] - ATR_MULT_SL * df.loc[i, 'ATR']) / df.loc[i, 'close']) - COMMISSION_FRAC)  \n",
    "\n",
    "        elif df.loc[i+2, 'high'] > (df.loc[i, 'close'] + ATR_MULT_TP * df.loc[i, 'ATR']):\n",
    "            up_outcomes.append(((df.loc[i, 'close'] + (ATR_MULT_TP * df.loc[i, 'ATR'])) / df.loc[i, 'close']) - COMMISSION_FRAC)\n",
    "\n",
    "        elif df.loc[i+2, 'low'] < (df.loc[i, 'close'] - ATR_MULT_SL * df.loc[i, 'ATR']):\n",
    "            up_outcomes.append(((df.loc[i, 'close'] - ATR_MULT_SL * df.loc[i, 'ATR']) / df.loc[i, 'close']) - COMMISSION_FRAC)\n",
    "\n",
    "        else:\n",
    "            up_outcomes.append((df.loc[i+2, 'close'] / df.loc[i, 'close']) - COMMISSION_FRAC)\n",
    "\n",
    "        up_outcomes_passive.append((df.loc[i+2, 'close'] / df.loc[i, 'close']) - (1.5 * COMMISSION_FRAC))\n",
    "\n",
    "up_outcomes.extend([float('NaN'), float('NaN')])\n",
    "up_outcomes_passive.extend([float('NaN'), float('NaN')])\n",
    "\n",
    "\n",
    "dn_outcomes = []\n",
    "dn_outcomes_passive = []\n",
    "\n",
    "# tp_frac = (df.loc[i, 'close'] / (df.loc[i, 'close'] - (ATR_MULT * df.loc[i, 'ATR']))) - COMMISSION_FRAC \n",
    "# sl_frac = (df.loc[i, 'close'] / (df.loc[i, 'high'] + ATR_MULT * df.loc[i, 'ATR'])) - COMMISSION_FRAC   \n",
    "\n",
    "for i in range(len(df)-2):\n",
    "\n",
    "    if bool(random.getrandbits(1)) == True:    \n",
    "\n",
    "        if df.loc[i+1, 'high'] > (df.loc[i, 'close'] + ATR_MULT_SL * df.loc[i, 'ATR']):\n",
    "            dn_outcomes.append((df.loc[i, 'close'] / (df.loc[i, 'close'] + ATR_MULT_SL * df.loc[i, 'ATR'])) - COMMISSION_FRAC) \n",
    "\n",
    "        elif df.loc[i+1, 'low'] < (df.loc[i, 'close'] - ATR_MULT_TP * df.loc[i, 'ATR']):    \n",
    "            dn_outcomes.append((df.loc[i, 'close'] / (df.loc[i, 'close'] - (ATR_MULT_TP * df.loc[i, 'ATR']))) - COMMISSION_FRAC) \n",
    "\n",
    "        elif df.loc[i+2, 'high'] > (df.loc[i, 'close'] + ATR_MULT_SL * df.loc[i, 'ATR']):\n",
    "            dn_outcomes.append((df.loc[i, 'close'] / (df.loc[i, 'close'] + ATR_MULT_SL * df.loc[i, 'ATR'])) - COMMISSION_FRAC) \n",
    "\n",
    "        elif df.loc[i+2, 'low'] < (df.loc[i, 'close'] - ATR_MULT_TP * df.loc[i, 'ATR']):\n",
    "            dn_outcomes.append((df.loc[i, 'close'] / (df.loc[i, 'close'] - (ATR_MULT_TP * df.loc[i, 'ATR']))) - COMMISSION_FRAC)\n",
    "\n",
    "        else:\n",
    "            dn_outcomes.append((df.loc[i, 'close'] / df.loc[i+2, 'close']) - COMMISSION_FRAC)\n",
    "\n",
    "        dn_outcomes_passive.append((df.loc[i, 'close'] / df.loc[i+2, 'close']) - (1.5 * COMMISSION_FRAC))\n",
    "\n",
    "    else:    \n",
    "\n",
    "        if df.loc[i+1, 'low'] < (df.loc[i, 'close'] - ATR_MULT_TP * df.loc[i, 'ATR']):    \n",
    "            dn_outcomes.append((df.loc[i, 'close'] / (df.loc[i, 'close'] - (ATR_MULT_TP * df.loc[i, 'ATR']))) - COMMISSION_FRAC) \n",
    "\n",
    "        elif df.loc[i+1, 'high'] > (df.loc[i, 'close'] + ATR_MULT_SL * df.loc[i, 'ATR']):\n",
    "            dn_outcomes.append((df.loc[i, 'close'] / (df.loc[i, 'close'] + ATR_MULT_SL * df.loc[i, 'ATR'])) - COMMISSION_FRAC) \n",
    "\n",
    "        elif df.loc[i+2, 'low'] < (df.loc[i, 'close'] - ATR_MULT_TP * df.loc[i, 'ATR']):\n",
    "            dn_outcomes.append((df.loc[i, 'close'] / (df.loc[i, 'close'] - (ATR_MULT_TP * df.loc[i, 'ATR']))) - COMMISSION_FRAC)\n",
    "\n",
    "        elif df.loc[i+2, 'high'] > (df.loc[i, 'close'] + ATR_MULT_SL * df.loc[i, 'ATR']):\n",
    "            dn_outcomes.append((df.loc[i, 'close'] / (df.loc[i, 'close'] + ATR_MULT_SL * df.loc[i, 'ATR'])) - COMMISSION_FRAC) \n",
    "\n",
    "        else:\n",
    "            dn_outcomes.append((df.loc[i, 'close'] / df.loc[i+2, 'close']) - COMMISSION_FRAC)\n",
    "\n",
    "        dn_outcomes_passive.append((df.loc[i, 'close'] / df.loc[i+2, 'close']) - (1.5 * COMMISSION_FRAC))\n",
    "\n",
    "dn_outcomes.extend([float('NaN'), float('NaN')])\n",
    "dn_outcomes_passive.extend([float('NaN'), float('NaN')])\n",
    "\n",
    "# # remove NaN rows to match main dataframe\n",
    "# for index in sorted(nan_indices, reverse=True):\n",
    "#     try:\n",
    "#         del up_outcomes[index]\n",
    "#         del dn_outcomes[index]\n",
    "#         del up_outcomes_passive[index]\n",
    "#         del dn_outcomes_passive[index]\n",
    "#     except:\n",
    "#         print('index out of range ', index)\n",
    "\n",
    "# print(len(up_outcomes))\n",
    "# print(len(dn_outcomes))\n",
    "# print(len(up_outcomes_passive))\n",
    "# print(len(dn_outcomes_passive))\n",
    "\n",
    "\n",
    "# filter results to match slicing of input tensors\n",
    "up_outcomes = [up_outcomes[i] for i in range(WINDOW, len(df), STEP)]\n",
    "dn_outcomes = [dn_outcomes[i] for i in range(WINDOW, len(df), STEP)]\n",
    "up_outcomes_passive = [up_outcomes_passive[i] for i in range(WINDOW, len(df), STEP)]\n",
    "dn_outcomes_passive = [dn_outcomes_passive[i] for i in range(WINDOW, len(df), STEP)]\n",
    "\n",
    "print(len(up_outcomes))\n",
    "print(len(dn_outcomes))\n",
    "print(len(up_outcomes_passive))\n",
    "print(len(dn_outcomes_passive))\n",
    "\n",
    "# save data\n",
    "with open(f'data/outcomes_{RUN_ID}{PL_ID}.pkl', 'wb') as file:\n",
    "    pickle.dump((up_outcomes, dn_outcomes, up_outcomes_passive, dn_outcomes_passive), file, protocol=4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 4. Separate catagorical dataframes and convert to one-hot encoding."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "      high_      low_    close_      vol_   close_f     vol_f      chh_      cll_        bk     d_sdev    c-c1    h-h1   l-l1        v-v1    hlc-hlc1      vwp-vwp1  rr_wvap_fast      rr_wvap          obv       acc_dis           mfv    tenkan     kijun  senkou_a  senkou_b    senkou    d_frama  rel_frama  frac_dim  d_frac_dim      ngd_3      ngd_4      ngd_6      ngd_8      ngd_11      ngd_16       ngd_22       ngd_32     drsi_7    drsi_14    drsi_21\n",
      "0 -0.004510 -0.012701 -0.012347 -0.461159  0.015008 -0.517846 -0.045379 -0.033702  1.536641 -14.265967    1.87  -16.22  132.9 -4987234.43  117.303333 -9.685257e+10   8370.694027  8370.636398 -22020735.69 -7.651111e+06 -2.413908e+06  0.994605  0.991600  0.000119  1.009113 -1.008995   1.870000   1.000000  0.676550   -0.309439   1.356111  28.734731  52.347200 -79.173469 -296.162754 -147.514770   163.464996  1563.706205  28.107646   8.320180   8.272585\n",
      "1 -0.007802 -0.012846 -0.011821 -0.382186  0.013615 -0.394398 -0.045267 -0.032725  1.380993   1.950352   -7.92  -40.34  -13.6   351272.15  -56.580000  6.713565e+09   8362.764955  8362.717235 -25052267.71 -9.408983e+06 -1.757872e+06  0.995994  0.990923  0.000119  1.008420 -1.008301  -3.946520   1.000475  1.151425    0.474875 -21.875830 -20.163856  43.223943 -43.771380 -215.778393  223.645640  -287.015957  1949.646241  26.693610   8.231494   8.194979\n",
      "2 -0.007670 -0.035685 -0.031554  0.710574 -0.009321  0.947037 -0.047631 -0.009874 -1.076812  -2.659984 -185.09  -17.37 -211.3  5582737.78 -290.366667  1.066069e+11   8177.623066  8177.578446 -33666537.51 -1.545742e+07 -6.048441e+06  1.014718  1.011962  0.000124  1.030668 -1.030544 -18.906348   1.020802  1.578398    0.426972 -44.844179 -61.974341 -32.181569 -22.305413 -110.351664  495.000850  -677.209838  2227.017657  20.882666   7.789607   7.726372\n",
      "3 -0.023704 -0.030384 -0.024286 -0.396493 -0.004530 -0.318473 -0.042391 -0.015431 -0.561038  -4.172543   45.01 -151.82   28.5 -5677747.52 -108.316667 -1.096345e+11   8222.647081  8222.606875 -30730015.23 -1.307531e+07  2.382113e+06  1.008179  1.007795  0.000123  1.014536 -1.014413 -12.514713   1.013694  1.543185   -0.035213  -3.727744 -24.078934 -79.165749 -14.582752   -9.616627  646.678859  -980.874143  2395.774031  34.970302  10.406715   9.950343\n",
      "4 -0.004557 -0.024222 -0.015429 -0.073422  0.002207  0.081646 -0.034694 -0.022937  0.196629  -1.812801   60.80  147.38   38.2  1726797.43  205.846667  3.401977e+10   8283.462135  8283.426508 -26066695.52 -1.356520e+07 -4.898881e+05  0.998347  1.001760  0.000121  1.007091 -1.006970  -5.183242   1.005630  1.543185    0.000000  52.390076  57.526015 -25.841362  12.869726   84.011138  692.676093 -1170.799129  2467.183458  40.492242  12.498963  12.079363 \n",
      " 50484\n",
      "   day_1  day_2  day_3  day_4  day_5  day_6  day_7  hour_0  hour_1  hour_2  hour_3  hour_4  hour_5  hour_6  hour_7  hour_8  hour_9  hour_10  hour_11  hour_12  hour_13  hour_14  hour_15  hour_16  hour_17  hour_18  hour_19  hour_20  hour_21  hour_22  hour_23  year_2018  year_2019  year_2020  year_2021  year_2022  year_2023  year_2024\n",
      "0  False  False  False  False   True  False  False   False   False   False   False   False   False   False   False   False   False    False     True    False    False    False    False    False    False    False    False    False    False    False    False       True      False      False      False      False      False      False\n",
      "1  False  False  False  False   True  False  False   False   False   False   False   False   False   False   False   False   False    False    False     True    False    False    False    False    False    False    False    False    False    False    False       True      False      False      False      False      False      False\n",
      "2  False  False  False  False   True  False  False   False   False   False   False   False   False   False   False   False   False    False    False    False     True    False    False    False    False    False    False    False    False    False    False       True      False      False      False      False      False      False\n",
      "3  False  False  False  False   True  False  False   False   False   False   False   False   False   False   False   False   False    False    False    False    False     True    False    False    False    False    False    False    False    False    False       True      False      False      False      False      False      False\n",
      "4  False  False  False  False   True  False  False   False   False   False   False   False   False   False   False   False   False    False    False    False    False    False     True    False    False    False    False    False    False    False    False       True      False      False      False      False      False      False \n",
      " 50484\n",
      "    down   flat     up\n",
      "0   True  False  False\n",
      "1  False   True  False\n",
      "2  False  False   True\n",
      "3  False   True  False\n",
      "4  False   True  False \n",
      " 50484\n"
     ]
    }
   ],
   "source": [
    "# create separate dataframe for one-hot encoded day/hour categories\n",
    "df_time = df[['day', 'hour', 'year']].copy()\n",
    "df_time = pd.get_dummies(df_time, columns=['day', 'hour', 'year'])\n",
    "\n",
    "# create separate dataframe for one-hot encoded target categories\n",
    "df_ylabels = df['target'].copy()\n",
    "df_ylabels = pd.get_dummies(df_ylabels, columns=['target'])\n",
    "\n",
    "# tidy up price data\n",
    "df = df.drop(columns=['unix', 'date', 'volume USD', 'volume', \n",
    "                      'open', 'high', 'low', 'close', \n",
    "                      f'SMA_{PERIOD}', f'SMA_{PERIOD_FAST}', \n",
    "                      f'vol SMA_{PERIOD}', f'vol SMA_{PERIOD_FAST}', \n",
    "                      'day', 'hour', 'year', 'target', 'ATR', 'open_',\n",
    "                      f'HH{PERIOD_FAST*2}', f'LL{PERIOD_FAST*2}', \n",
    "                      ], axis=1)\n",
    "\n",
    "# check dataframes\n",
    "print(df.head(), '\\n', len(df))\n",
    "print(df_time.head(), '\\n', len(df_time))\n",
    "print(df_ylabels.head(), '\\n', len(df_ylabels))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5. Scale standard deviations of selected columns then check overall balance of data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "COUNT: 50484\n",
      " down     6845\n",
      "flat    36135\n",
      "up       7504\n",
      "dtype: int64\n",
      "MEANS:\n",
      " high_           2.301352e-01\n",
      "low_           -1.889741e-01\n",
      "close_          2.900061e-02\n",
      "vol_           -6.843780e-18\n",
      "close_f         1.703022e-02\n",
      "vol_f           4.593239e-02\n",
      "chh_            7.656589e-17\n",
      "cll_           -7.825485e-17\n",
      "bk              4.016762e-02\n",
      "d_sdev          9.841987e-05\n",
      "c-c1            4.121209e-03\n",
      "h-h1            4.586821e-03\n",
      "l-l1            3.793377e-03\n",
      "v-v1           -7.195549e-05\n",
      "hlc-hlc1        5.064199e-03\n",
      "vwp-vwp1       -6.513259e-05\n",
      "rr_wvap_fast   -6.755814e-17\n",
      "rr_wvap        -1.261085e-16\n",
      "obv             3.377907e-16\n",
      "acc_dis         1.846589e-16\n",
      "mfv            -4.644622e-18\n",
      "tenkan          6.161302e-15\n",
      "kijun           8.948216e-15\n",
      "senkou_a       -3.377907e-16\n",
      "senkou_b       -4.258133e-15\n",
      "senkou         -1.090219e-15\n",
      "d_frama         5.066860e-18\n",
      "rel_frama      -1.176789e-14\n",
      "frac_dim        1.481071e-15\n",
      "d_frac_dim     -3.615416e-18\n",
      "ngd_3          -2.744549e-18\n",
      "ngd_4           1.140044e-17\n",
      "ngd_6          -7.037306e-18\n",
      "ngd_8          -2.251938e-18\n",
      "ngd_11          8.444767e-19\n",
      "ngd_16         -7.741037e-18\n",
      "ngd_22          1.252641e-17\n",
      "ngd_32         -1.604506e-17\n",
      "drsi_7         -9.922074e-16\n",
      "drsi_14        -9.344487e-16\n",
      "drsi_21         6.391985e-16\n",
      "dtype: float64\n",
      "STDEVS:\n",
      " high_           0.978668\n",
      "low_            1.015410\n",
      "close_          1.000000\n",
      "vol_            1.000000\n",
      "close_f         1.000000\n",
      "vol_f           1.000000\n",
      "chh_            1.000000\n",
      "cll_            1.000000\n",
      "bk              1.000000\n",
      "d_sdev          1.000000\n",
      "c-c1            1.000000\n",
      "h-h1            1.000000\n",
      "l-l1            1.000000\n",
      "v-v1            1.000000\n",
      "hlc-hlc1        1.000000\n",
      "vwp-vwp1        1.000000\n",
      "rr_wvap_fast    1.000000\n",
      "rr_wvap         1.000000\n",
      "obv             1.000000\n",
      "acc_dis         1.000000\n",
      "mfv             1.000000\n",
      "tenkan          1.000000\n",
      "kijun           1.000000\n",
      "senkou_a        1.000000\n",
      "senkou_b        1.000000\n",
      "senkou          1.000000\n",
      "d_frama         1.000000\n",
      "rel_frama       1.000000\n",
      "frac_dim        1.000000\n",
      "d_frac_dim      1.000000\n",
      "ngd_3           1.000000\n",
      "ngd_4           1.000000\n",
      "ngd_6           1.000000\n",
      "ngd_8           1.000000\n",
      "ngd_11          1.000000\n",
      "ngd_16          1.000000\n",
      "ngd_22          1.000000\n",
      "ngd_32          1.000000\n",
      "drsi_7          1.000000\n",
      "drsi_14         1.000000\n",
      "drsi_21         1.000000\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "import pickle\n",
    "\n",
    "if PERCENT_CHANGE == True:\n",
    "    df = df.pct_change()\n",
    "    print(df.head())\n",
    "    print(df.tail())\n",
    "    df = df.dropna(axis=0)\n",
    "    df.replace([np.inf, -np.inf], np.nan, inplace=True)\n",
    "\n",
    "# scale column stdev with stdev(close_)\n",
    "df['high_'] = df['high_'] / df['close_'].std()\n",
    "df['low_'] = df['low_'] / df['close_'].std()\n",
    "df['close_'] = df['close_'] / df['close_'].std()\n",
    "\n",
    "# scale column stdev to unity\n",
    "df['vol_'] = (df['vol_'] - df['vol_'].mean()) / df['vol_'].std()   \n",
    "df['close_f'] = df['close_f'] / df['close_f'].std()\n",
    "df['vol_f'] = df['vol_f'] / df['vol_f'].std()\n",
    "df['chh_'] = (df['chh_'] - df['chh_'].mean()) / df['chh_'].std()\n",
    "df['cll_'] = (df['cll_'] - df['cll_'].mean()) / df['cll_'].std()\n",
    "df['bk'] = df['bk'] / df['bk'].std()\n",
    "df['d_sdev'] = df['d_sdev'] / df['d_sdev'].std()\n",
    "df['rr_wvap_fast'] = (df['rr_wvap_fast'] - df['rr_wvap_fast'].mean()) / df['rr_wvap_fast'].std()\n",
    "df['rr_wvap'] = (df['rr_wvap'] - df['rr_wvap'].mean()) / df['rr_wvap'].std()\n",
    "df['vwp-vwp1'] = df['vwp-vwp1'] / df['vwp-vwp1'].std()\n",
    "\n",
    "df['c-c1'] = df['c-c1'] / df['c-c1'].std()\n",
    "df['h-h1'] = df['h-h1'] / df['h-h1'].std()      # are these just replicating first layer's filters?\n",
    "df['l-l1'] = df['l-l1'] / df['l-l1'].std()\n",
    "df['v-v1'] = df['v-v1'] / df['v-v1'].std()\n",
    "df['hlc-hlc1'] = df['hlc-hlc1'] / df['hlc-hlc1'].std()\n",
    "\n",
    "df['tenkan'] = (df['tenkan'] - df['tenkan'].mean()) / df['tenkan'].std()    \n",
    "df['kijun'] = (df['kijun'] - df['kijun'].mean()) / df['kijun'].std()\n",
    "df['senkou'] = (df['senkou'] - df['senkou'].mean()) / df['senkou'].std()\n",
    "df['senkou_a'] = (df['senkou_a'] - df['senkou_a'].mean()) / df['senkou_a'].std()\n",
    "df['senkou_b'] = (df['senkou_b'] - df['senkou_b'].mean()) / df['senkou_b'].std()\n",
    "\n",
    "df['acc_dis'] = (df['acc_dis'] - df['acc_dis'].mean()) / df['acc_dis'].std()\n",
    "df['obv'] = (df['obv'] - df['obv'].mean()) / df['obv'].std()\n",
    "df['mfv'] = (df['mfv'] - df['mfv'].mean()) / df['mfv'].std()\n",
    "\n",
    "df['d_frama'] = (df['d_frama'] - df['d_frama'].mean()) / df['d_frama'].std()    \n",
    "df['rel_frama'] = (df['rel_frama'] - df['rel_frama'].mean()) / df['rel_frama'].std()\n",
    "df['frac_dim'] = (df['frac_dim'] - df['frac_dim'].mean()) / df['frac_dim'].std()\n",
    "df['d_frac_dim'] = (df['d_frac_dim'] - df['d_frac_dim'].mean()) / df['d_frac_dim'].std()\n",
    "\n",
    "df['drsi_7'] = (df['drsi_7'] - df['drsi_7'].mean()) / df['drsi_7'].std()\n",
    "df['drsi_14'] = (df['drsi_14'] - df['drsi_14'].mean()) / df['drsi_14'].std()\n",
    "df['drsi_21'] = (df['drsi_21'] - df['drsi_21'].mean()) / df['drsi_21'].std()\n",
    "\n",
    "df['ngd_3'] = (df['ngd_3'] - df['ngd_3'].mean()) / df['ngd_3'].std()\n",
    "df['ngd_4'] = (df['ngd_4'] - df['ngd_4'].mean()) / df['ngd_4'].std()\n",
    "df['ngd_6'] = (df['ngd_6'] - df['ngd_6'].mean()) / df['ngd_6'].std()\n",
    "df['ngd_8'] = (df['ngd_8'] - df['ngd_8'].mean()) / df['ngd_8'].std()\n",
    "df['ngd_11'] = (df['ngd_11'] - df['ngd_11'].mean()) / df['ngd_11'].std()\n",
    "df['ngd_16'] = (df['ngd_16'] - df['ngd_16'].mean()) / df['ngd_16'].std()\n",
    "df['ngd_22'] = (df['ngd_22'] - df['ngd_22'].mean()) / df['ngd_22'].std()\n",
    "df['ngd_32'] = (df['ngd_32'] - df['ngd_32'].mean()) / df['ngd_32'].std()\n",
    "\n",
    "# check data\n",
    "print(f'COUNT: {len(df)}\\n', df_ylabels.sum(axis=0))\n",
    "print('MEANS:\\n', df.mean())\n",
    "print('STDEVS:\\n', df.std())\n",
    "\n",
    "# save data\n",
    "with open(f'data/dataframes_{RUN_ID}.pkl', 'wb') as file:\n",
    "    pickle.dump((df, df_time, df_ylabels), file, protocol=4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare Tensors\n",
    "#### Slices dataframe into many discrete classification problems."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "price_series (25232, 20, 41)\n",
      "time_categories (25232, 38)\n",
      "target_categories (25232, 3)\n",
      "up_targets (25232, 2)\n",
      "dn_targets (25232, 2)\n"
     ]
    }
   ],
   "source": [
    "# from numba import jit\n",
    "\n",
    "\n",
    "# @jit\n",
    "def prepare_tensors(df, df_time, df_ylabels, window, step): \n",
    "\n",
    "    # create numpy arrays to receive data\n",
    "    price_series = np.zeros(shape=(window, len(df.columns)))\n",
    "    time_categories = np.zeros(shape=(1, len(df_time.columns)))\n",
    "    target_categories = np.zeros(shape=(1, 3))\n",
    "\n",
    "    # iterate through price dataframe concatenating discrete arrays of size 'window', and spacing 'step'\n",
    "    batch_size = (len(df)-window) // step\n",
    "    for i in range(batch_size): \n",
    "        arr = df.iloc[[(i*step)+j for j in range(window)], [k for k in range(len(df.columns))]].to_numpy()\n",
    "        price_series = np.concatenate((price_series, arr))\n",
    "\n",
    "    # iterate through categorical dataframes concatenating data relating to bottom row of each price window\n",
    "    for i in range(batch_size): \n",
    "        arr = df_time.iloc[[(i*step) + window], : ].to_numpy()\n",
    "        time_categories = np.concatenate((time_categories, arr))\n",
    "        \n",
    "        arr = df_ylabels.iloc[[(i*step) + window], : ].to_numpy()\n",
    "        target_categories = np.concatenate((target_categories, arr))\n",
    "\n",
    "    # reshape arrays\n",
    "    price_series = np.reshape(price_series, (batch_size+1, window, len(df.columns)))\n",
    "    time_categories = np.reshape(time_categories, (batch_size+1, len(df_time.columns)))\n",
    "    target_categories = np.reshape(target_categories, (batch_size+1, 3))\n",
    "\n",
    "    # delete intial 'zeros' array elements\n",
    "    price_series = np.delete(price_series, 0, axis=0)\n",
    "    time_categories = np.delete(time_categories, 0, axis=0)\n",
    "    target_categories = np.delete(target_categories, 0, axis=0)\n",
    "\n",
    "    # check arrays\n",
    "    print('price_series', price_series.shape)\n",
    "    print('time_categories', time_categories.shape)\n",
    "    print('target_categories', target_categories.shape)\n",
    "\n",
    "    # binarize target categories, index(1) == positive class\n",
    "    up_targets = target_categories[ : , [1, 0]]\n",
    "    dn_targets = target_categories[ : , [1, 2]]\n",
    "\n",
    "    for element in up_targets:\n",
    "        if element[0] == 0 and element[1] == 0:\n",
    "            element[0] = 1\n",
    "\n",
    "    for element in dn_targets:\n",
    "        if element[0] == 0 and element[1] == 0:\n",
    "            element[0] = 1\n",
    "\n",
    "    # check arrays\n",
    "    print('up_targets', up_targets.shape)\n",
    "    print('dn_targets', dn_targets.shape)\n",
    "\n",
    "    return price_series, time_categories, up_targets, dn_targets\n",
    "\n",
    "\n",
    "# with open(f'data/dataframes_{RUN_ID}.pkl', 'rb') as file:\n",
    "#     df, df_time, df_ylabels = pickle.load(file)\n",
    "\n",
    "price_series, time_categories, up_targets, dn_targets = prepare_tensors(df, df_time, df_ylabels, WINDOW, STEP)\n",
    "\n",
    "# save data\n",
    "with open(f'data/tensors_{RUN_ID}.pkl', 'wb') as file:\n",
    "    pickle.dump((price_series, time_categories, up_targets, dn_targets), file, protocol=4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Notes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "## add commission level to 'flat' calculation\n",
    "## frama\n",
    "## are all inputs normally distributed? - what happens if not... inverse Fisher Transform \n",
    "## noise mask on inputs\n",
    "## oc2 removes noise and half a bar lag in transforms?\n",
    "## rolling st_dev for normalisation to avoid amplitude increase over time - or rate of change? - does sq_rt of series improve accuracy? - or weights recent data by accident?\n",
    "## ...log(price) best?  pd.pct_change() np.log(series)\n",
    "## replace math functions with numpy for numba\n",
    "## risk adjusted outcomes"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}