-
Notifications
You must be signed in to change notification settings - Fork 0
/
eda_feature_importance.py
126 lines (95 loc) · 4.32 KB
/
eda_feature_importance.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
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
import random
import warnings
from collections import Counter
warnings.filterwarnings('ignore')
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import MinMaxScaler
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.ensemble import AdaBoostClassifier
import xgboost as xgb
print('\nLoading files ...')
filename = "train.csv"
train = pd.read_csv(filename)
train = train.reset_index(drop=True)
#sum ind bins
train['ps_ind_sum_bin'] = np.zeros(train.shape[0])
dcol = [c for c in train.columns if c not in ['id','target']]
for c in dcol:
if '_bin' in c and 'ps_ind_' in c: #standard arithmetic
train['ps_ind_sum_bin'] += train[c]
#sum calc bins
train['ps_calc_sum_bin'] = np.zeros(train.shape[0])
dcol = [c for c in train.columns if c not in ['id','target']]
for c in dcol:
if '_bin' in c and 'ps_calc_' in c: #standard arithmetic
train['ps_calc_sum_bin'] += train[c]
print(c)
train['ps_sqrt(car_15)*reg_03'] = train['ps_reg_03']*np.sqrt(train['ps_car_15'])
train['ps_sqrt(car_15)*reg_02'] = train['ps_reg_02']*np.sqrt(train['ps_car_15'])
train['ps_sqrt(car_13)*reg_02'] = np.sqrt(train['ps_car_13']) * train['ps_reg_02']
train['ps_sqrt(sum_reg)'] = np.sqrt(1+train['ps_reg_03']+train['ps_reg_02']+train['ps_reg_01'])
#train['ps_car_sqrt(13+15)'] = np.sqrt(train['ps_car_13']+train['ps_car_15'])
train['ps_car_sqrt(13+15)/reg_01'] = np.sqrt(train['ps_car_13']+train['ps_car_15'])*np.sqrt(train['ps_reg_01'])
# train['ps_reg_F'] = train['ps_reg_03'].apply(lambda x: recon(x)[0])
# train['ps_reg_F'],_ = scale_data(train['ps_reg_F'].reshape(-1, 1))
# train['ps_CALC'] = train['ps_reg_F']/train['ps_car_13']
# #train['ps_CALC'] = np.sqrt(2+train['ps_reg_F'] + train['ps_reg_03'])
zeros_like = (train['target'] == 0)*1
ones_like = (train['target'] == 1)*1.5
samples_w = zeros_like+ones_like
#### FEATURES
def recon(reg):
integer = int(np.round((40*reg)**2)) # gives 2060 for our example
for f in range(28):
if (integer - f) % 27 == 0:
F = f
M = (integer - F)//27
return F, M
def scale_data(X, scaler=None):
if not scaler:
scaler = MinMaxScaler()#feature_range=(-1, 1))
scaler.fit(X)
X = scaler.transform(X)
return X, scaler
fig, axis = plt.subplots(1,1)
# train['ps_reg_F'] = train['ps_reg_03'].apply(lambda x: recon(x)[0])
# train['ps_reg_F'],_ = scale_data(train['ps_reg_F'].reshape(-1, 1))
# train['ps_CALC'] = np.sqrt(2+train['ps_reg_F'] + train['ps_reg_03']) #np.sqrt(14+train['ps_car_12']*train['ps_car_13']*train['ps_car_14']*train['ps_car_15'])
#### TRAINING
zeros_like = (train['target'] == 0)*1
ones_like = (train['target'] == 1)*1.5
samples_w = zeros_like+ones_like
print(samples_w)
algorithm = AdaBoostClassifier(n_estimators = 100,learning_rate = 0.75)
algorithm2 = RandomForestClassifier(n_estimators=100, max_depth=8, min_samples_leaf=4, max_features=0.2, n_jobs=-1, random_state=0)
algorithm3 = GradientBoostingClassifier(n_estimators=100, max_depth=8, min_samples_leaf=4, max_features=0.2, random_state=0)
algorithm.fit(train.drop(['id', 'target'],axis=1), train.target, sample_weight=samples_w)
algorithm2.fit(train.drop(['id', 'target'],axis=1), train.target, sample_weight=samples_w)
algorithm3.fit(train.drop(['id', 'target'],axis=1), train.target, sample_weight=samples_w)
print("----- Training Done -----")
# Scatter plot
features = train.drop(['id', 'target'],axis=1).columns.values
importances = (algorithm.feature_importances_ + algorithm2.feature_importances_ + algorithm3.feature_importances_)/3
indices = np.argsort(importances)
print("importances--")
print(importances.sum())
# df = pd.DataFrame(importances, columns=['feature', 'fscore'])
# plt.figure()
# df.plot()
# df.plot(kind='barh', x='feature', y='fscore', legend=False, figsize=(10, 25))
# plt.gcf().savefig('feature_importance/GBM-1.png')
print(importances)
print("")
print(indices)
plt.title('Feature Importances')
plt.barh(range(len(indices)), importances[indices], color='b', align='center')
plt.yticks(range(len(indices)), features[indices]) ## removed [indices]
plt.xlabel('Relative Importance')
plt.subplots_adjust(top=0.9, bottom=0.10, left=0.10, right=0.95, hspace=0.65,
wspace=0.7)
plt.show()