-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathml-libs.py
171 lines (128 loc) · 6.01 KB
/
ml-libs.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
import numpy as np
import pandas as pd
from scipy.sparse import csc_matrix, eye, diags
from scipy.sparse.linalg import spsolve
from xgboost import XGBRegressor
from sklearn.model_selection import GridSearchCV
from sklearn.preprocessing import Normalizer
from sklearn.decomposition import PCA
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
import seaborn as sns
import os
from sklearn.metrics import mean_squared_error, r2_score, mean_absolute_error
from joblib import dump, load
from sklearn.neural_network import MLPRegressor, MLPClassifier
import pickle as pk
def read_data(filepath):
return pd.read_csv(filepath)
def divide_data(data):
X = data.drop(['Cr','Mn','Mo','Ni'], axis = 1)
y_Cr = data['Cr']
y_Mn = data['Mn']
y_Mo = data['Mo']
y_Ni = data['Ni']
y = {'Cr': y_Cr, 'Mn': y_Mn, 'Mo': y_Mo, 'Ni': y_Ni}
return X, y
def filter_noise(X, treshold):
X_limited = X.iloc[:,X.columns.astype('float')<220]
X = X.assign(noise_level = X_limited.apply(max, axis=1))
return X[X['noise_level'] < treshold].drop(['noise_level'],axis=1)
def limit_wavelength(X,w_min = 225, w_max = 940):
X_limited = X.loc[:,(X.columns.astype('float')>w_min) & (X.columns.astype('float')<w_max)]
return X_limited
def update_dependent_variable(X, data_training):
y_Cr = data_training[data_training.index.isin(X.index)]['Cr']
y_Mn = data_training[data_training.index.isin(X.index)]['Mn']
y_Mo = data_training[data_training.index.isin(X.index)]['Mo']
y_Ni = data_training[data_training.index.isin(X.index)]['Ni']
return y_Cr, y_Mn, y_Mo, y_Ni
def tune_xgb_model(X_train, y_train):
xgb = XGBRegressor(random_state=123)
parameters = {'nthread': [4], # when use hyperthread, xgboost may become slower
'objective': ['reg:squarederror'],
'learning_rate': [.03, 0.05, .07], # so called `eta` value
'max_depth': [5, 6, 7],
'min_child_weight': [4],
'subsample': [0.7],
'colsample_bytree': [0.7],
'n_estimators': [500]}
xgb_grid = GridSearchCV(xgb,
parameters,
cv=5,
n_jobs=5,
verbose=True)
xgb_grid.fit(X_train, y_train)
print('Best score:', xgb_grid.best_score_)
print('Best parameters:', xgb_grid.best_params_)
return xgb_grid.best_estimator_
def preprocessed_data_pca(X, data_training, element='Mn',filter=False):
if filter:
X = filter_noise(X=X, treshold=0.05)
X_avg_limited = limit_wavelength(X, w_min=225, w_max=940)
y_Cr, y_Mn, y_Mo, y_Ni = update_dependent_variable(X_avg_limited, data_training)
y = {'Cr': y_Cr, 'Mn': y_Mn, 'Mo': y_Mo, 'Ni': y_Ni}
pca = PCA(n_components=15)
X_normalized = Normalizer(norm='l2').fit_transform(X_avg_limited)
X_train, X_test, y_train, y_test = train_test_split(X_normalized, y[element], test_size=0.33, random_state=42)
pca.fit(X_train)
projected_train = pca.transform(X_train)
return projected_train, y_train, pca, X_test, y_test
def plot(x, y, element, r2, mae):
reg_plot = sns.regplot(x=x, y=y, ci=95)
reg_plot.set_xlabel('True', fontsize=20)
reg_plot.set_ylabel('Prediction', fontsize=20)
reg_plot.set_title(element + r' $R^{2}$: ' + str(r2) + ' MAE: ' + str(mae), fontsize=20)
reg_plot.grid()
plt.savefig(os.path.join('report',element,element+'.png'))
plt.clf()
def report(y_test, y_pred, element, report_df):
r2 = np.round(r2_score(y_test, y_pred), 3)
mae = np.round(mean_absolute_error(y_test, y_pred), 3)
report_df[element] = [r2, mae]
report_df['score'] = ['R2', 'MAE']
report_df = report_df.set_index('score')
report_df.to_csv(os.path.join('report','xgboost_report.csv'))
return r2, mae
def train():
report_df_pca = pd.DataFrame()
data_training = read_data(os.path.join('data', 'train_dataset.csv'))
X_training, y = divide_data(data_training)
for element in ['Cr', 'Mn', 'Mo', 'Ni']:
print(element)
projected_train, y_train, pca, X_test, y_test = preprocessed_data_pca(X=X_training, data_training=data_training,
element=element)
pk.dump(pca, open(os.path.join("saved-pca","pca.pkl"), "wb"))
xgb_tuned_model = tune_xgb_model(projected_train, y_train)
dump(xgb_tuned_model, os.path.join('saved-models', element, "xgboost_" + element + ".joblib"))
projected_test = pca.transform(X_test)
y_pred = xgb_tuned_model.predict(projected_test)
r2, mae = report(y_test, y_pred, element, report_df_pca)
x = np.array(y_test)
y = np.array(y_pred)
plot(x, y, element, r2, mae)
def test(pca,element):
test_df = pd.read_csv(os.path.join('data','test_dataset.csv'))
targets = test_df['target_name'].unique()
predictions_df = pd.DataFrame()
for target in targets:
predictions = []
print(target, element)
target_df = test_df.loc[test_df['target_name'] == target].drop(['target_name'],axis=1)
target_df_limited = limit_wavelength(target_df)
target_normalized = Normalizer(norm='l2').fit_transform(target_df_limited)
target_normalized_df = pd.DataFrame(target_normalized)
for index, row in target_normalized_df.iterrows():
xgb_model = load(os.path.join('saved-models',element,'xgboost_'+element+'.joblib'))
predictions.append(xgb_model.predict(pca.transform(row.values.reshape(1, -1))))
predictions_df[target] = [np.average(predictions), np.std(predictions)]
predictions_df['score'] = ['pred', 'unc']
predictions_df = predictions_df.set_index('score')
predictions_df.to_csv(os.path.join('test',element,'test_'+element+'.csv'))
def test_all():
pca = pk.load(open(os.path.join("saved-pca", "pca.pkl"), 'rb'))
for element in ['Cr', 'Mn', 'Mo', 'Ni']:
test(pca,element)
if __name__ == "__main__":
train()
test_all()