src/Predicting_HeartDisease.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "MSk_HeartDisease.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "mount_file_id": "180qxsUM0r3kSa5syr27sIUfhOl0CtoWh",
      "authorship_tag": "ABX9TyPTZ23kfFTOxZ3Vr/swa3Id",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/maciejskorski/ml_examples/blob/master/Predicting_HeartDisease.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SjtmM__l0-2I"
      },
      "source": [
        "# Summary\n",
        " **What are most important risk factors for heart diseases?** We answer this question using\n",
        "the Cleveland UC Irvine challenge, which asks to predict a heart disease based on 13 predictors (both numeric and categorical):\n",
        "age, sex, chest pain symptoms, blood press,\tcholesterol, blood sugar, rest ecg,\tmax heart rate,\texercise angina,\tpeak and slope on ecg, blocked vessels and\tthallium test. This setup has been discussed many times as a toy or research problem, but this analysis prioritizes *robustness* to mitigate the selection bias (rarely addressed in other discussions).  "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oaSX6ybLKyus",
        "cellView": "form"
      },
      "source": [
        "#@title Preprocessing \n",
        "#@markdown  6 rows with missing data was skipped, leaving 297 partient records. The original names were changed to be more descriptive and different disease variants were merged into one (hdisease).   \n",
        "import pandas as pd\n",
        "_=!curl \"https://archive.ics.uci.edu/ml/machine-learning-databases/heart-disease/processed.cleveland.data\" -o \"data.csv\"\n",
        "df=pd.read_csv('data.csv',header=None,skiprows=0)\n",
        "df.columns = ['age','sex','chestpain','rest_bloodpress','cholesterol','bloodsugar','restecg','maxhrate','exangina','oldpeak','slope','blockedvessels','thallium','hdisease']\n",
        "# convert target (different deases are encoded)\n",
        "df['hdisease'] = df['hdisease']>0\n",
        "# drop few rows with NA values\n",
        "df=df.drop(df.index[(df=='?').any(1)],axis=0)\n",
        "# force types\n",
        "df['blockedvessels'] = df['blockedvessels'].astype('float')\n",
        "df['thallium'] = df['thallium'].astype('float')\n",
        "# give meaningful labels\n",
        "#encoder = {1:'typ_angin',2:'atyp_angin',3:'non-angin',4:'asymptom'}\n",
        "#df['chestpain'] = df['chestpain'].astype('int').apply(encoder.get)\n",
        "#encoder = {1:'up',2:'flat',3:'down'}\n",
        "#df['slope'] = df['slope'].astype('int').apply(encoder.get)\n",
        "#encoder = {3:'norm',6:'fixed',7:'reversible'}\n",
        "#df['thallium'] = df['thallium'].astype('int').apply(encoder.get)\n",
        "#encoder = {0:'norm',1:'abnorm',2:'hypertrophy_evidence'}\n",
        "#df['restecg'] = df['restecg'].astype('int').apply(encoder.get)\n",
        "# show rows\n",
        "#df.head()\n",
        "#print(df.head().to_markdown())\n",
        "#from IPython.display import display, HTML\n",
        "#display(HTML(df.head().to_html()))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8sS5egZLbIGH"
      },
      "source": [
        "## Feature Selection\n",
        "\n",
        "**Model**: Random Forest (parameters adjusted to avoid overly complex trees) was used to estimate the importance of features. Rather than fitting it on the whole data, this solution cross-validates it to minimize the selection bias (interval estimates for importances)\n",
        "\n",
        "**Conclusions** thallium test, blocked vessels, chest pains, exercise angina and certain ecg patterns are major contributting factors. These detailed diagnostics and symptoms are ranked higher than more general (age, sex) or causal (sugar level) predictors. Some of the ecg metrics have low importance, maybe they are overprovisioned?"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 350
        },
        "id": "yItg28WFw9I_",
        "outputId": "73e6e49b-85ea-4048-cbd2-d4b6edbd29b0"
      },
      "source": [
        "#@title Random Forest Model (Cross-Validated)\n",
        "from sklearn.metrics import make_scorer,roc_auc_score\n",
        "from sklearn.feature_selection import RFECV,SelectFromModel\n",
        "from sklearn.model_selection import cross_val_score,StratifiedKFold,cross_validate\n",
        "from sklearn.ensemble import RandomForestClassifier\n",
        "from sklearn.linear_model import LogisticRegression\n",
        "from matplotlib import pyplot as plt\n",
        "\n",
        "y = df['hdisease']\n",
        "X = df[df.columns[:-1]]\n",
        "model = RandomForestClassifier(n_estimators=200,max_depth=5,min_samples_leaf=5,random_state=13)\n",
        "cv = StratifiedKFold(5,shuffle=True,random_state=13)\n",
        "out = cross_validate(model,X,y,scoring='f1',cv=cv,return_estimator=True)\n",
        "\n",
        "ranks = []\n",
        "for m in out['estimator']:\n",
        "  ranks.append(pd.Series(m.feature_importances_,index=X.columns))\n",
        "ranks = pd.DataFrame(ranks).T\n",
        "ranks['score'] = ranks.mean(1)\n",
        "ranks = ranks.sort_values('score',ascending=True)\n",
        "fig,ax = plt.subplots(figsize=(10,5))\n",
        "ranks.T.boxplot(vert=False,figsize=(14,10),patch_artist=True,ax=ax)\n",
        "ax.set_title('Feature Importance (cross-validated random forest)')\n",
        "ax.set_xlabel('Importance (with std bars)')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7yQO4q54bYmO"
      },
      "source": [
        "## Generalized Linear Model (BIC Selection)\n",
        "\n",
        "**Model** <emph>Recursive Feature Elimination</emph> is used with Logistic Regression as the building block and Bayesian Information Criterion as the overfitting measure; in every round the least significant feature with p-value>0.10 is deleted. The model was *optimized by introducing non-linearity* in form of \"breaking points\" at \"extreme\" levels of cholesterol, blood presure, age, heart rate etc.\n",
        "\n",
        "**Conclusions** The strongest impact (by sigificance) comes as before from blocked vessels, chest pain (4=atypical), heart rate, thalllium. However this time age, sex and cholesterol are important. This may due to the protective estrogen impact, but why this was not seen in the previous model?"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "cellView": "form",
        "id": "xiSVJGDs3GNw",
        "outputId": "3ae59d85-043c-46ca-c3b4-bb7d87390c6c"
      },
      "source": [
        "#@title Logistic Regression\n",
        "X_onehot = X.copy(deep=True)\n",
        "\n",
        "cat_vars = ['sex','chestpain','restecg','bloodsugar','exangina','slope','thallium']\n",
        "vals = []\n",
        "names = []\n",
        "for c in cat_vars:\n",
        "  val = pd.get_dummies(X_onehot[c])\n",
        "  val.columns = list(map(str,val.columns))\n",
        "  X_onehot = X_onehot.drop(c,axis=1)\n",
        "  val = val[list(val.columns)[1:]]\n",
        "  val.columns = list(map(lambda x: c+str(x),val.columns))\n",
        "  vals.append(val)\n",
        "\n",
        "X_onehot = pd.concat([X_onehot,pd.concat(vals,axis=1)],axis=1)\n",
        "X_onehot['age>55'] = (X_onehot['age']>57).astype('float')\n",
        "X_onehot['cholesterol>275'] = (X_onehot['cholesterol']>275).astype('float')\n",
        "X_onehot['rest_bloodpress>150'] = (X_onehot['rest_bloodpress'] > 150).astype('float')\n",
        "X_onehot['maxhrate>187'] = (X_onehot['maxhrate'] > 170).astype('float')\n",
        "\n",
        "import statsmodels.api as sm\n",
        "import statsmodels.formula.api  as smf\n",
        "from statsmodels.tools import eval_measures\n",
        "\n",
        "ftrs = list(X_onehot.columns)\n",
        "outs = []\n",
        "\n",
        "end = False\n",
        "while not end:\n",
        "  model = sm.GLM(exog=X_onehot[ftrs], endog=y.astype(int), family=sm.families.Binomial())\n",
        "  model = model.fit()\n",
        "  worst = model.pvalues.argmax(-1)\n",
        "  if model.pvalues[worst] > 0.10:\n",
        "    outs.append((ftrs[worst],eval_measures.bic(model.llf,model.nobs,model.df_model)))\n",
        "    ftrs.remove(ftrs[worst])\n",
        "  else:\n",
        "    end = True\n",
        "\n",
        "_=print(model.summary().tables[1])\n",
        "print('start BIC=%s,end BIC=%s'%(outs[0][1],outs[-1][1]))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "===================================================================================\n",
            "                      coef    std err          z      P>|z|      [0.025      0.975]\n",
            "-----------------------------------------------------------------------------------\n",
            "age                -0.0850      0.028     -3.013      0.003      -0.140      -0.030\n",
            "rest_bloodpress     0.0209      0.010      2.002      0.045       0.000       0.041\n",
            "maxhrate           -0.0241      0.008     -3.002      0.003      -0.040      -0.008\n",
            "oldpeak             0.3447      0.199      1.735      0.083      -0.045       0.734\n",
            "blockedvessels      1.2773      0.269      4.746      0.000       0.750       1.805\n",
            "sex1.0              1.5233      0.480      3.175      0.001       0.583       2.464\n",
            "chestpain2.0        1.1261      0.607      1.854      0.064      -0.065       2.317\n",
            "chestpain4.0        1.8913      0.438      4.316      0.000       1.032       2.750\n",
            "exangina1.0         0.7620      0.446      1.710      0.087      -0.112       1.636\n",
            "slope2.0            1.0398      0.418      2.490      0.013       0.221       1.858\n",
            "thallium7.0         1.4390      0.405      3.549      0.000       0.644       2.234\n",
            "age>55              1.4961      0.581      2.573      0.010       0.357       2.636\n",
            "cholesterol>275     0.7882      0.464      1.698      0.089      -0.121       1.698\n",
            "===================================================================================\n",
            "start BIC=303.3493621912014,end BIC=261.21451062554183\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        },
        "cellView": "form",
        "id": "ZQwv-GBsoD5s",
        "outputId": "0d55c532-376b-42f8-972a-315a4fac0a1f"
      },
      "source": [
        "#@title Diagnostic Plots\n",
        "from matplotlib import pyplot as plt\n",
        "\n",
        "infl = model.get_influence()\n",
        "fig,axs = plt.subplots(1,2,figsize=(10,4))\n",
        "infl.plot_index(y_var='cooks', threshold=3*4/300, ax=axs[0])\n",
        "infl.plot_index(y_var='resid', threshold=3, ax=axs[1])\n",
        "fig.tight_layout(pad=1.0)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nuAufyyly6N4"
      },
      "source": [
        "## Model Comparison\n",
        "\n",
        "Both models give similar insights and are tested against overfitting. The simpler linear model actually achieves a sligthly better F1 score ($0.83\\pm 0.06$ as opposed to $0.8\\pm 0.05$, on cross-validation folds). The linear model can be extended into a hierarchical model which is aware of intra-cluster effects (e.g. effects common for group of patients) and be beneficial for tasks such as sample stratification. As a side remark, the literature reports that getting close to 0.9  is very hard.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bT5SjVIky80F",
        "outputId": "0ee34fea-e8fa-4aba-9019-ff177d401f21"
      },
      "source": [
        "model = RandomForestClassifier(n_estimators=200,max_depth=5,min_samples_leaf=10,random_state=13)\n",
        "out = cross_val_score(model,X,y,scoring='f1',cv=cv)\n",
        "print(out.mean().round(2),out.std().round(2))\n",
        "\n",
        "model = LogisticRegression(max_iter=1000)\n",
        "out = cross_val_score(model,X_onehot[ftrs],y,scoring='f1',cv=cv)\n",
        "print(out.mean().round(2),out.std().round(2))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.8 0.05\n",
            "0.83 0.05\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0Vh0l4eLM72L"
      },
      "source": [
        "## Extras"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "cellView": "form",
        "id": "Xya-czxG6QvO"
      },
      "source": [
        "#@title Export summary PDF\n",
        "%%capture\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive', force_remount=True)\n",
        "#_=!sudo apt-get install texlive-xetex texlive-fonts-recommended texlive-generic-recommended\n",
        "!jupyter nbconvert --no-input --template drive/MyDrive/Colab\\ Notebooks/better_template --to PDF drive/MyDrive/Colab\\ Notebooks/HeartDisease.ipynb --"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7Vzk3YMOeBds"
      },
      "source": [
        "### Improving Fit\n",
        "\n",
        "The fit is quite good as also confirmed by the diagnostic plots. The anomalies can be skipped to improve the fit (but does not change the insights); for example the person 218 is healthy despite several risk factors: 64-old, male, high cholesterol and blocked main vessels.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 795
        },
        "cellView": "form",
        "id": "cXSGQK_zv5je",
        "outputId": "8aab25ab-ceba-40df-a883-9ddb70602747"
      },
      "source": [
        "#@title Improving Logistic Regression\n",
        "X_onehot = X.copy(deep=True)\n",
        "\n",
        "# see where are anomalies\n",
        "blacklist = abs(infl.summary_frame()['standard_resid'])>3\n",
        "df.loc[blacklist,]\n",
        "\n",
        "vals = []\n",
        "names = []\n",
        "for c in cat_vars:\n",
        "  val = pd.get_dummies(X_onehot[c])\n",
        "  val.columns = list(map(str,val.columns))\n",
        "  X_onehot = X_onehot.drop(c,axis=1)\n",
        "  val = val[list(val.columns)[1:]]\n",
        "  val.columns = list(map(lambda x: c+str(x),val.columns))\n",
        "  vals.append(val)\n",
        "\n",
        "X_onehot = pd.concat([X_onehot,pd.concat(vals,axis=1)],axis=1)\n",
        "X_onehot['age>55'] = (X_onehot['age']>57).astype('float')\n",
        "X_onehot['cholesterol>275'] = (X_onehot['cholesterol']>275).astype('float')\n",
        "X_onehot['rest_bloodpress>150'] = (X_onehot['rest_bloodpress'] > 150).astype('float')\n",
        "X_onehot['maxhrate>187'] = (X_onehot['maxhrate'] > 170).astype('float')\n",
        "\n",
        "import statsmodels.api as sm\n",
        "import statsmodels.formula.api  as smf\n",
        "from statsmodels.tools import eval_measures\n",
        "\n",
        "ftrs = list(X_onehot.columns)\n",
        "outs = []\n",
        "\n",
        "end = False\n",
        "while not end:\n",
        "  model = sm.GLM(exog=X_onehot.loc[~blacklist][ftrs], endog=y[~blacklist].astype(int), family=sm.families.Binomial())\n",
        "  model = model.fit()\n",
        "  worst = model.pvalues.argmax(-1)\n",
        "  if model.pvalues[worst] > 0.2:\n",
        "    outs.append((ftrs[worst],eval_measures.bic(model.llf,model.nobs,model.df_model)))\n",
        "    ftrs.remove(ftrs[worst])\n",
        "  else:\n",
        "    end = True\n",
        "\n",
        "print(model.summary())\n",
        "print('start BIC=%s,end BIC=%s'%(outs[0][1],outs[-1][1]))\n",
        "\n",
        "infl = model.get_influence()\n",
        "\n",
        "fig,axs = plt.subplots(1,2,figsize=(10,4))\n",
        "infl.plot_index(y_var='cooks', threshold=3*4/300, ax=axs[0])\n",
        "infl.plot_index(y_var='resid', threshold=3, ax=axs[1])\n",
        "fig.tight_layout(pad=1.0)\n",
        "\n",
        "model = LogisticRegression(max_iter=1000)\n",
        "out = cross_val_score(model,X_onehot[ftrs],y,scoring='f1',cv=5)\n",
        "print(out.mean(),out.std())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "                 Generalized Linear Model Regression Results                  \n",
            "==============================================================================\n",
            "Dep. Variable:               hdisease   No. Observations:                  289\n",
            "Model:                            GLM   Df Residuals:                      277\n",
            "Model Family:                Binomial   Df Model:                           11\n",
            "Link Function:                  logit   Scale:                          1.0000\n",
            "Method:                          IRLS   Log-Likelihood:                -61.652\n",
            "Date:                Sun, 25 Apr 2021   Deviance:                       123.30\n",
            "Time:                        09:59:41   Pearson chi2:                     167.\n",
            "No. Iterations:                     8                                         \n",
            "Covariance Type:            nonrobust                                         \n",
            "===================================================================================\n",
            "                      coef    std err          z      P>|z|      [0.025      0.975]\n",
            "-----------------------------------------------------------------------------------\n",
            "age                -0.1413      0.038     -3.732      0.000      -0.216      -0.067\n",
            "rest_bloodpress     0.0353      0.013      2.676      0.007       0.009       0.061\n",
            "maxhrate           -0.0381      0.011     -3.464      0.001      -0.060      -0.017\n",
            "oldpeak             0.5623      0.267      2.105      0.035       0.039       1.086\n",
            "blockedvessels      2.6141      0.458      5.712      0.000       1.717       3.511\n",
            "sex1.0              2.1872      0.671      3.258      0.001       0.872       3.503\n",
            "chestpain2.0        2.5214      0.833      3.026      0.002       0.888       4.155\n",
            "chestpain4.0        3.2176      0.657      4.898      0.000       1.930       4.505\n",
            "exangina1.0         0.8393      0.551      1.523      0.128      -0.241       1.919\n",
            "slope2.0            2.0913      0.595      3.514      0.000       0.925       3.258\n",
            "thallium7.0         2.5744      0.563      4.571      0.000       1.470       3.678\n",
            "age>55              2.3342      0.769      3.035      0.002       0.827       3.842\n",
            "===================================================================================\n",
            "start BIC=236.67781113698356,end BIC=189.70047038995267\n",
            "0.8408952267775798 0.04437827383453979\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5JCv0J-zRTjL"
      },
      "source": [
        "## Explanatory Analysis :-)\n",
        "\n",
        "We will have a look and see if the features discriminate the response (information-theoretically)."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 142
        },
        "id": "AE1Gmxa0RVIY",
        "outputId": "4f5f7a4f-8f77-499f-b0d6-f5effa124462",
        "cellView": "form"
      },
      "source": [
        "#@title Gender Impact \n",
        "#@markdown Being a female looks like an advantage. Probably this \"estrogen\" conjecture.\n",
        "df.groupby('sex')[['hdisease']].mean()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>hdisease</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>sex</th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0.0</th>\n",
              "      <td>0.260417</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1.0</th>\n",
              "      <td>0.557214</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "     hdisease\n",
              "sex          \n",
              "0.0  0.260417\n",
              "1.0  0.557214"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 17
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "id": "WEb3XsStR5Wg",
        "outputId": "b0a2e8c8-615d-46d0-96b8-0e0e2b1d546d",
        "cellView": "form"
      },
      "source": [
        "#@title Chest Pain Impact \n",
        "#@markdown The risk goes up with pain of type 4, which is \"asymptomatic\" wrt anginal symptoms.\n",
        "df.groupby('chestpain')[['hdisease']].mean()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>hdisease</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>chestpain</th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>1.0</th>\n",
              "      <td>0.304348</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2.0</th>\n",
              "      <td>0.183673</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3.0</th>\n",
              "      <td>0.216867</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4.0</th>\n",
              "      <td>0.725352</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "           hdisease\n",
              "chestpain          \n",
              "1.0        0.304348\n",
              "2.0        0.183673\n",
              "3.0        0.216867\n",
              "4.0        0.725352"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 142
        },
        "id": "9REnPIdJSach",
        "outputId": "0380f9f9-ac33-488b-936d-e56f4420e011",
        "cellView": "form"
      },
      "source": [
        "#@title Blood Sugar Impact \n",
        "#@markdown The sugar alone revels no information on the target. I suppose it is cousal: indirectly triggers other precise metrics.\n",
        "df.groupby('bloodsugar')[['hdisease']].mean()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>hdisease</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>bloodsugar</th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0.0</th>\n",
              "      <td>0.460630</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1.0</th>\n",
              "      <td>0.465116</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "            hdisease\n",
              "bloodsugar          \n",
              "0.0         0.460630\n",
              "1.0         0.465116"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 173
        },
        "id": "OraSUpHOSxFh",
        "outputId": "15d8a4b9-4dad-472b-a455-4509ce6be7d1",
        "cellView": "form"
      },
      "source": [
        "#@title Rest ECG Impact \n",
        "#@markdown Value 1 appears with higher risk. It encodes \"ST-T abnormality\", so this makes sense.\n",
        "df.groupby('restecg')[['hdisease']].mean()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>hdisease</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>restecg</th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0.0</th>\n",
              "      <td>0.374150</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1.0</th>\n",
              "      <td>0.750000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2.0</th>\n",
              "      <td>0.541096</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "         hdisease\n",
              "restecg          \n",
              "0.0      0.374150\n",
              "1.0      0.750000\n",
              "2.0      0.541096"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 25
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 142
        },
        "id": "AMwGDDcNTpo2",
        "outputId": "c23aada3-0e2e-40ff-d864-4ce854b164f0",
        "cellView": "form"
      },
      "source": [
        "#@title Exercise-induced Angina Impact \n",
        "#@markdown This is a known risk factor, visible in the data.\n",
        "df.groupby('exangina')[['hdisease']].mean()\n",
        "\t"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>hdisease</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>exangina</th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0.0</th>\n",
              "      <td>0.315000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1.0</th>\n",
              "      <td>0.762887</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "          hdisease\n",
              "exangina          \n",
              "0.0       0.315000\n",
              "1.0       0.762887"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 29
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "id": "uym4WnFCUCcq",
        "outputId": "3f8885ee-b458-45c8-b2c5-aa910caf1815",
        "cellView": "form"
      },
      "source": [
        "#@title Blocked Major Vessels Impact \n",
        "#@markdown The more, the worse! Also known to the literature.\n",
        "df.groupby('blockedvessels')[['hdisease']].mean()\n",
        "\t"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>hdisease</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>blockedvessels</th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0.0</th>\n",
              "      <td>0.258621</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1.0</th>\n",
              "      <td>0.676923</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2.0</th>\n",
              "      <td>0.815789</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3.0</th>\n",
              "      <td>0.850000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                hdisease\n",
              "blockedvessels          \n",
              "0.0             0.258621\n",
              "1.0             0.676923\n",
              "2.0             0.815789\n",
              "3.0             0.850000"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 38
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 173
        },
        "id": "NErlUgR7Vt-t",
        "outputId": "dabb5c8d-79b6-4e44-ac04-c489bd2d8718",
        "cellView": "form"
      },
      "source": [
        "#@title Thallium Test Impact \n",
        "#@markdown The value of 3, which is the \"normal reading\", decreases the risk (again makes sense)\n",
        "df.groupby('thallium')[['hdisease']].mean()\n",
        "\t"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>hdisease</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>thallium</th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>3.0</th>\n",
              "      <td>0.225610</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>6.0</th>\n",
              "      <td>0.666667</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7.0</th>\n",
              "      <td>0.765217</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "          hdisease\n",
              "thallium          \n",
              "3.0       0.225610\n",
              "6.0       0.666667\n",
              "7.0       0.765217"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 40
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 285
        },
        "id": "MdAjya7OWka_",
        "outputId": "a9db68e1-ebab-43d1-873e-771307cef85e",
        "cellView": "form"
      },
      "source": [
        "#@title Cholesterol\n",
        "#@markdown There is some impact from  very high cholesterol levels.\n",
        "mask = df['hdisease']==True\n",
        "df[mask]['cholesterol'].plot.kde(label='disease')\n",
        "df[~mask]['cholesterol'].plot.kde(label='no disease')\n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7ff9c7c539d0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 61
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 285
        },
        "id": "lgqlLJ-tZfzv",
        "outputId": "d4fcef78-b34e-49e0-e360-3d36a0751d2e",
        "cellView": "form"
      },
      "source": [
        "#@title Max Heart Rate\n",
        "#@markdown There is some impact but negative, from  the high heart rate, don't understand what is is this. I think this may have something to do with young age?\n",
        "mask = df['hdisease']==True\n",
        "df[mask]['maxhrate'].plot.kde(label='disease')\n",
        "df[~mask]['maxhrate'].plot.kde(label='no disease')\n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7ff9c3b78d50>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 62
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "cellView": "form",
        "id": "evq93ZiwaJHF"
      },
      "source": [
        "#@title Export Full PDF\n",
        "%%capture\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive', force_remount=True)\n",
        "#_=!sudo apt-get install texlive-xetex texlive-fonts-recommended texlive-generic-recommended\n",
        "!jupyter nbconvert --template drive/MyDrive/Colab\\ Notebooks/better_template --to PDF drive/MyDrive/Colab\\ Notebooks/HeartDisease.ipynb --"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_VjA2OQXfB4I",
        "outputId": "cbb0bc49-e1e0-463c-9d54-59f3cf69c9ab"
      },
      "source": [
        "X_onehot.loc[218]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "age                     64.0\n",
              "rest_bloodpress        130.0\n",
              "cholesterol            303.0\n",
              "maxhrate               122.0\n",
              "oldpeak                  2.0\n",
              "blockedvessels           2.0\n",
              "sex1.0                   0.0\n",
              "chestpain2.0             0.0\n",
              "chestpain3.0             0.0\n",
              "chestpain4.0             1.0\n",
              "restecg1.0               0.0\n",
              "restecg2.0               0.0\n",
              "bloodsugar1.0            0.0\n",
              "exangina1.0              0.0\n",
              "slope2.0                 1.0\n",
              "slope3.0                 0.0\n",
              "thallium6.0              0.0\n",
              "thallium7.0              0.0\n",
              "age>55                   1.0\n",
              "cholesterol>275          1.0\n",
              "rest_bloodpress>150      0.0\n",
              "maxhrate>187             0.0\n",
              "Name: 218, dtype: float64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 25
        }
      ]
    }
  ]
}