-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtraining_model_one_cardio.py
252 lines (227 loc) · 12.5 KB
/
training_model_one_cardio.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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
import numpy as np
import pandas as pd
from sklearn.externals import joblib
from sklearn.ensemble import GradientBoostingClassifier, AdaBoostClassifier, RandomForestClassifier
from sklearn.model_selection import train_test_split
from sklearn.neural_network import MLPClassifier
from sklearn.metrics import log_loss
from processing_data import LoadFile
data = pd.DataFrame(LoadFile("ml5/train.csv"))
# Нормализация данных
'''
data_n = data[['age', 'height', 'weight', 'ap_hi', 'ap_lo',
'cholesterol', 'gluc', 'bmi', 'ap_hi_n', 'ap_lo_n']]
data_n = (data_n - data_n.mean()) / data_n.std()
data[['age', 'height', 'weight', 'ap_hi', 'ap_lo', 'cholesterol', 'gluc', 'bmi', 'ap_hi_n', 'ap_lo_n']
] = data_n[['age', 'height', 'weight', 'ap_hi', 'ap_lo', 'cholesterol', 'gluc', 'bmi', 'ap_hi_n', 'ap_lo_n']]
'''
# age;gender;height;weight;ap_hi;ap_lo;cholesterol;gluc;smoke;alco;active;cardio
# Предсказание курение алкоголь активность
#,'weight_o','weight_nfg_o','weight_nfg_o_с','weight_o_c'
'''
data_n = pd.DataFrame(StandardScaler().fit_transform(data[['age', 'height', 'weight', 'ap_hi', 'ap_lo',
'cholesterol', 'gluc', 'bmi', 'ap_hi_n', 'ap_lo_n','weight_o','weight_nfg_o','weight_nfg_o_с','weight_o_c']]))
data[['age', 'height', 'weight', 'ap_hi', 'ap_lo', 'cholesterol', 'gluc', 'bmi', 'ap_hi_n', 'ap_lo_n', 'weight_o', 'weight_nfg_o', 'weight_nfg_o_с', 'weight_o_c']
] = data_n
X_train.loc[(X_train['predict']>0.1) & (X_train['smoke']==1),'predict']=X_train[(X_train.predict>0.1) & (X_train.smoke==1)]['predict']+0.2
X_test.loc[(X_test['predict']>0.1) & (X_test['smoke']==1),'predict']=X_test[(X_test.predict>0.1) & (X_test.smoke==1)]['predict']+0.2
X_train.loc[(X_train['predict']>0.1) & (X_train['alco']==1),'predict']=X_train[(X_train.predict>0.1) & (X_train.alco==1)]['predict']+0.2
X_test.loc[(X_test['predict']>0.1) & (X_test['alco']==1),'predict']=X_test[(X_test.predict>0.1) & (X_test.alco==1)]['predict']+0.2
X_train.loc[(X_train['predict']>0.1) & (X_train['bmi_r_1']==1),'predict']=X_train[(X_train.predict>0.1) & (X_train.bmi_r_1==1)]['predict']+0.2
X_test.loc[(X_test['predict']>0.1) & (X_test['bmi_r_1']==1),'predict']=X_test[(X_test.predict>0.1) & (X_test.bmi_r_1==1)]['predict']+0.2
X_train.loc[(X_train['predict']>0.1) & (X_train['bmi_r_3']==1),'predict']=X_train[(X_train.predict>0.1) & (X_train.bmi_r_3==1)]['predict']+0.3
X_test.loc[(X_test['predict']>0.1) & (X_test['bmi_r_3']==1),'predict']=X_test[(X_test.predict>0.1) & (X_test.bmi_r_3==1)]['predict']+0.3
X_train.loc[(X_train['predict']>0.1) & (X_train['bmi_r_4']==1),'predict']=X_train[(X_train.predict>0.1) & (X_train.bmi_r_4==1)]['predict']+0.4
X_test.loc[(X_test['predict']>0.1) & (X_test['bmi_r_4']==1),'predict']=X_test[(X_test.predict>0.1) & (X_test.bmi_r_4==1)]['predict']+0.4
0.253346938776 0.265761904762 X = data.drop(['cardio','id','weight_o','weight_nfg_o','weight_nfg_o_с','weight_o_c','alco'], axis=1) # Выбрасываем столбец 'class'.
'''
#print(X.columns)
# Предсказание курения
print("start gbt")
#abc = ensemble.AdaBoostClassifier(n_estimators=300, random_state=264)
#gbt = ensemble.GradientBoostingClassifier(n_estimators=300, random_state=264)
'''
0.240673469388 0.267523809524 err_sum 0.248728571429(n_estimators=300, random_state=264,learning_rate = 0.3)
0.254408163265 0.265238095238 err_sum 0.257657142857(n_estimators=300, random_state=264,learning_rate = 0.1)
0.0591836734694 0.282428571429 err_sum 0.126157142857(n_estimators=300, random_state=264,max_depth = 10)
0.253979591837 0.26480952381 err_sum 0.257228571429(n_estimators=300, random_state=264,min_samples_leaf = 3)
0.25387755102 0.264761904762 err_sum 0.257142857143(n_estimators=300, random_state=264,min_samples_leaf = 5)
0.254265306122 0.264428571429 err_sum 0.257314285714(n_estimators=300, random_state=264,min_samples_leaf = 5, subsample = 0.5)
start to nite
,criterion = 'mae'
'''
X = data.drop(['cardio','active','id','weight_o','weight_o_c','alco','bmi_r_4','bmi_n_7','bmi_r_1','bmi_n_2','bmi_n_1'], axis=1) # Выбрасываем столбец 'class'.
Y = data['cardio']
#print(X.describe())
'''
#gbt = ensemble.GradientBoostingClassifier(n_estimators=55, random_state=264,min_samples_leaf = 5, subsample = 0.5, verbose=0)
gender 0 0.257255812494 0.263050003661 err_sum 0.258994069844 gbt = MLPClassifier(alpha=0.0, random_state = 0)
gender 0 0.261052367356 0.262025038436 err_sum 0.26134416868 gbt = MLPClassifier(alpha=0.0, random_state = 0, hidden_layer_sizes = 50, verbose = 1)
'''
best_err = 10
best_rnd = 0
runs = 40
data_predict = data.drop(['cardio','active','id','weight_o','weight_o_c','alco','bmi_r_4','bmi_n_7','bmi_r_1','bmi_n_2','bmi_n_1'], axis=1) # Выбрасываем столбец 'class'.
data_p_l = LoadFile("ml5/test.csv")
data_p_x = data_p_l.drop(['active','id','weight_o','weight_o_c','alco','bmi_r_4','bmi_n_7','bmi_r_1','bmi_n_2','bmi_n_1'], axis=1) # Выбрасываем столбец 'class'.
data_p = data_p_l.drop(['active','id','weight_o','weight_o_c','alco','bmi_r_4','bmi_n_7','bmi_r_1','bmi_n_2','bmi_n_1'], axis=1) # Выбрасываем столбец 'class'.
for i in range(runs):
X_train, X_test, Y_train, Y_test = train_test_split(
X, Y, test_size=0.3, random_state=i)
gbt = MLPClassifier(alpha=0.0, random_state = 5, activation = 'relu', hidden_layer_sizes=(50,), verbose = 0)
clf4 = gbt.fit(X_train, Y_train)
err_train = np.mean(Y_train != gbt.predict(X_train))
err_test = np.mean(Y_test != gbt.predict(X_test))
err_sum = np.mean(Y != gbt.predict(X))
data_predict["MLP_"+str(i)] = pd.DataFrame(gbt.predict_proba(X)).drop(0,axis=1)
data_p["MLP_"+str(i)] = pd.DataFrame(gbt.predict_proba(data_p_x)).drop(0,axis=1)
#joblib.dump(gbt, "training_models/cardio.pkl", compress=1)
l_los = log_loss(Y,gbt.predict_proba(X)[:, 1])
if(err_test<best_err):
best_err=err_test
best_rnd = i
print("random_state = ", i, err_train, err_test, 'err_sum', err_sum, "log_loss =",l_los)
#print("gbt score %s" % clf4.score(X_train, Y_train))
print("Best MLP rnd", best_rnd, "err", best_err)
best_err = 10
best_rnd = 0
for i in range(runs):
X_train, X_test, Y_train, Y_test = train_test_split(
X, Y, test_size=0.3, random_state=i)
gbt = GradientBoostingClassifier(n_estimators=55, random_state=264,min_samples_leaf = 5, subsample = 0.5, verbose=0)
clf4 = gbt.fit(X_train, Y_train)
err_train = np.mean(Y_train != gbt.predict(X_train))
err_test = np.mean(Y_test != gbt.predict(X_test))
err_sum = np.mean(Y != gbt.predict(X))
data_predict["GBT_"+str(i)] = pd.DataFrame(gbt.predict_proba(X)).drop(0,axis=1)
data_p["GBT_"+str(i)] = pd.DataFrame(gbt.predict_proba(data_p_x)).drop(0,axis=1)
#joblib.dump(gbt, "training_models/cardio.pkl", compress=1)
l_los = log_loss(Y,gbt.predict_proba(X)[:, 1])
if(err_test<best_err):
best_err=err_test
best_rnd = i
print("random_state = ", i, err_train, err_test, 'err_sum', err_sum, "log_loss =",l_los)
print("Best GBT rnd", best_rnd, "err", best_err)
best_err = 10
best_rnd = 0
for i in range(runs):
X_train, X_test, Y_train, Y_test = train_test_split(
X, Y, test_size=0.3, random_state=i)
gbt = AdaBoostClassifier(n_estimators=300, random_state=264)
clf4 = gbt.fit(X_train, Y_train)
err_train = np.mean(Y_train != gbt.predict(X_train))
err_test = np.mean(Y_test != gbt.predict(X_test))
err_sum = np.mean(Y != gbt.predict(X))
data_predict["ABT_"+str(i)] = pd.DataFrame(gbt.predict_proba(X)).drop(0,axis=1)
data_p["ABT_"+str(i)] = pd.DataFrame(gbt.predict_proba(data_p_x)).drop(0,axis=1)
#joblib.dump(gbt, "training_models/cardio.pkl", compress=1)
l_los = log_loss(Y,gbt.predict_proba(X)[:, 1])
if(err_test<best_err):
best_err=err_test
best_rnd = i
print("random_state = ", i, err_train, err_test, 'err_sum', err_sum, "log_loss =",l_los)
print("Best ABC rnd", best_rnd, "err", best_err)
#
best_err = 10
best_rnd = 0
for i in range(runs):
X_train, X_test, Y_train, Y_test = train_test_split(
X, Y, test_size=0.3, random_state=i)
gbt = RandomForestClassifier(max_depth=10, n_estimators=700, max_features=1, random_state=i)
clf4 = gbt.fit(X_train, Y_train)
err_train = np.mean(Y_train != gbt.predict(X_train))
err_test = np.mean(Y_test != gbt.predict(X_test))
err_sum = np.mean(Y != gbt.predict(X))
data_predict["RFC_"+str(i)] = pd.DataFrame(gbt.predict_proba(X)).drop(0,axis=1)
data_p["RFC_"+str(i)] = pd.DataFrame(gbt.predict_proba(data_p_x)).drop(0,axis=1)
#joblib.dump(gbt, "training_models/cardio.pkl", compress=1)
l_los = log_loss(Y,gbt.predict_proba(X)[:, 1])
if(err_test<best_err):
best_err=err_test
best_rnd = i
print("random_state = ", i, err_train, err_test, 'err_sum', err_sum, "log_loss =",l_los)
print("Best RFC rnd", best_rnd, "err", best_err)
#data_predict = data_predict.drop(['cardio'],axis=1)
'''best_err = 10
best_rnd = 0
best_rnd_model = 0
best_layer = 0
for j_m in range(runs):
X_train, X_test, Y_train, Y_test = train_test_split(
data_predict, Y, test_size=0.3, random_state=j_m)
for j in range(10, 200, 10):
gbt = MLPClassifier(alpha=0.0, random_state=j_m,
activation='relu', hidden_layer_sizes=(j,), verbose=0)
clf4 = gbt.fit(X_train, Y_train)
err_train = np.mean(Y_train != gbt.predict(X_train))
err_test = np.mean(Y_test != gbt.predict(X_test))
err_sum = np.mean(Y != gbt.predict(data_predict))
if(err_test<best_err):
best_err=err_test
best_rnd = j_m
best_layer = j
joblib.dump(gbt, "training_models/cardio_"+str(best_rnd)+"_"+str(best_layer)+".pkl", compress=1)
print("random_state =", j_m, err_train, err_test, 'err_sum', err_sum, "layer", j)
print("Best final rnd", best_rnd, "err", best_err, "best layer", best_layer)'''
X_train, X_test, Y_train, Y_test = train_test_split(
data_predict, Y, test_size=0.3, random_state=14)
gbt = MLPClassifier(alpha=0.0, random_state=14,
activation='relu', hidden_layer_sizes=(180,), verbose=0)
clf4 = gbt.fit(X_train, Y_train)
err_train = np.mean(Y_train != gbt.predict(X_train))
err_test = np.mean(Y_test != gbt.predict(X_test))
err_sum = np.mean(Y != gbt.predict(data_predict))
print("random_state = 14", err_train, err_test, 'err_sum', err_sum, "layer = 180")
data_p["predict"] = pd.DataFrame(gbt.predict_proba(data_p)).drop(0,axis=1)
data_p['predict'].to_csv("result/test_predict1.csv", sep=';', index=False)
data_p.to_csv("result/test_predict.csv", sep=';', index=False)
l_los = log_loss(Y,gbt.predict_proba(data_predict)[:, 1])
print("log loss", l_los)
print("File is save")
'''
whith smoke
Best MLP rnd 20 err 0.25819047619
Best GBT rnd 20 err 0.258285714286
Best ABC rnd 14 err 0.26180952381
Best RFC rnd 20 err 0.264714285714
Best final rnd 20 err 0.254047619048 best layer 130
whith smoke alco
Best MLP rnd 20 err 0.25819047619
Best GBT rnd 20 err 0.258285714286
Best ABC rnd 14 err 0.26180952381
Best RFC rnd 20 err 0.264714285714
Best final rnd 20 err 0.2540476 layer 130
whith smoke active alco
Best MLP rnd 33 err 0.259238095238
Best GBT rnd 20 err 0.259619047619
Best ABC rnd 20 err 0.260714285714
Best RFC rnd 20 err 0.265142857
whithout smoke, alco, active
Best MLP rnd 14 err 0.25819047619
Best GBT rnd 20 err 0.258476190476
Best ABC rnd 20 err 0.26219047619
Best RFC rnd 20 err 0.263619047619
Best final rnd 20 err 0.253761904762 best layer 70
data_predict = pd.DataFrame(data[['cardio', 'smoke', 'alco', 'active']])
Best MLP rnd 14 err 0.25819047619
Best GBT rnd 20 err 0.258476190476
Best ABC rnd 20 err 0.26219047619
Best RFC rnd 20 err 0.263619047619
Best final rnd 20 err 0.253952380952 best layer 10
#data_predict = data.drop(['cardio','smoke','alco','active','id','weight_o','weight_nfg_o','weight_nfg_o_с','weight_o_c','alco','bmi_r_4','bmi_n_7','bmi_r_1','bmi_n_2','bmi_n_1'], axis=1) # Выбрасываем столбец 'class'.
Best MLP rnd 14 err 0.25819047619
Best GBT rnd 20 err 0.258476190476
Best ABC rnd 20 err 0.26219047619
Best RFC rnd 20 err 0.263619047619
Best final rnd 14 err 0.217571428571 best layer 180
'''
'''print()
feature_names = X.columns
importances = gbt.feature_importances_
indices = np.argsort(importances)[::-1]
print("Feature importances:")
for f, idx in enumerate(indices):
print("{:2d}. feature '{:5s}' ({:.4f})".format(f + 1, feature_names[idx], importances[idx]))'''
#0.265285714286 0.265238095238 0.265238095238
#0.254959183673 0.265714285714 0.254408163265 0.265238095238
#0.260326530612 0.255714285714 e