-
Notifications
You must be signed in to change notification settings - Fork 32
/
Copy pathCPSC_hybrid.py
104 lines (82 loc) · 4.13 KB
/
CPSC_hybrid.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
# -*- coding: utf-8 -*-
"""
Created on Wed Dec 26 15:43:18 2018
@author: Winham
# CPSC_hybrid.py: 使用xgboost混合深度学习网络和人工特征，得到最后结果
"""
import os
import numpy as np
import xgboost as xgb
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix
from CPSC_config import Config
import CPSC_utils as utils
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
config = Config()
config.MODEL_PATH = 'E:/CPSC_Scheme/Net_models/'
config.MAN_FEATURE_PATH = 'E:/CPSC_Scheme/Man_features/'
records_name = np.array(os.listdir(config.DATA_PATH))
records_label = np.load(config.REVISED_LABEL) - 1
class_num = len(np.unique(records_label))
train_val_records, _, train_val_labels, test_labels = train_test_split(
records_name, records_label, test_size=0.2, random_state=config.RANDOM_STATE)
train_records, val_records, train_labels, val_labels = train_test_split(
train_val_records, train_val_labels, test_size=0.2, random_state=config.RANDOM_STATE)
_, train_labels = utils.oversample_balance(train_records, train_labels, config.RANDOM_STATE)
_, val_labels = utils.oversample_balance(val_records, val_labels, config.RANDOM_STATE)
# 载入之前保存的网络输出概率以及人工特征 -------------------------------------------------------------------------------
pred_nnet_r = np.load(config.MODEL_PATH + 'pred_nnet_r.npy')
pred_nnet_v = np.load(config.MODEL_PATH + 'pred_nnet_v.npy')
pred_nnet_t = np.load(config.MODEL_PATH + 'pred_nnet_t.npy')
man_features_r = np.load(config.MAN_FEATURE_PATH + 'man_features_r.npy')
man_features_v = np.load(config.MAN_FEATURE_PATH + 'man_features_v.npy')
man_features_t = np.load(config.MAN_FEATURE_PATH + 'man_features_t.npy')
pred_r = np.concatenate((pred_nnet_r, man_features_r), axis=1)
pred_v = np.concatenate((pred_nnet_v, man_features_v), axis=1)
pred_t = np.concatenate((pred_nnet_t, man_features_t), axis=1)
lb_r = train_labels
lb_v = val_labels
lb_t = test_labels
# 训练xgboost （仅根据验证集表现调参）----------------------------------------------------------------------------------
dtrain = xgb.DMatrix(pred_r, label=lb_r)
dval = xgb.DMatrix(pred_v, label=lb_v)
dtest = xgb.DMatrix(pred_t, label=lb_t)
param = [('max_depth', 10), ('objective', 'multi:softmax'),
('eval_metric', 'merror'), ('subsample', 0.5),
('eta', 0.01), ('num_class', 9), ('min_child_weight', 1.0),
('gamma', 0.2), ('colsample_bytree', 0.7),
('lambda', 20), ('max_delta_step', 7)
]
watchlist = [(dtrain, 'train'), (dval, 'test')]
num_round = 25
bst = xgb.train(param, dtrain, num_round, watchlist)
# 评估在过采样验证集，原始验证集，以及测试集上的性能 -------------------------------------------------------------------
pred = bst.predict(dval)
Conf_Matv = confusion_matrix(lb_v, pred)
print('\nResult for oversampling val_set:--------------------\n')
F1s_val = []
for j in range(Conf_Matv.shape[0]):
f1vt = 2*Conf_Matv[j][j]/(np.sum(Conf_Matv[j, :])+np.sum(Conf_Matv[:, j]))
print('| F1-'+config.CLASS_NAME[j]+':'+str(f1vt)+' |')
F1s_val.append(f1vt)
print('\nF1-mean: ' + str(np.mean(F1s_val)))
val_records_raw, val_records_ind = np.unique(val_records, return_index=True)
lb_v_raw = lb_v[val_records_ind]
pred_raw = pred[val_records_ind]
Conf_Matv_raw = confusion_matrix(lb_v_raw,pred_raw)
print('\nResult for raw val_set:--------------------\n')
F1s_val_raw = []
for j in range(Conf_Matv_raw.shape[0]):
f1vt = 2*Conf_Matv_raw[j][j]/(np.sum(Conf_Matv_raw[j, :])+np.sum(Conf_Matv_raw[:, j]))
print('| F1-'+config.CLASS_NAME[j]+':'+str(f1vt)+' |')
F1s_val_raw.append(f1vt)
print('\nF1-mean: ' + str(np.mean(F1s_val_raw)))
pred = bst.predict(dtest)
Conf_Matt = confusion_matrix(lb_t, pred)
print('\nResult for test_set:--------------------\n')
F1s_test = []
for j in range(Conf_Matt.shape[0]):
f1tt = 2*Conf_Matt[j][j]/(np.sum(Conf_Matt[j, :])+np.sum(Conf_Matt[:, j]))
print('| F1-'+config.CLASS_NAME[j]+':'+str(f1tt)+' |')
F1s_test.append(f1tt)
print('\nF1-mean: ' + str(np.mean(F1s_test)))