-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathxgb_tsa.py
176 lines (149 loc) · 6.91 KB
/
xgb_tsa.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
# /usr/bin/python
# -*- coding: utf-8 -*-
import sys
import zipfile
import pandas as pd
from scipy import sparse
from sklearn.preprocessing import OneHotEncoder
from sklearn.linear_model import LogisticRegression
# sys.path.append('C:\\Users\\DAWIN\\Documents\\Tencent Files\\694596886\\FileRecv\\XGBoost包及教程\\xgboost-master\\wrapper')
import xgboost as xgb
from sklearn.cross_validation import train_test_split
import time
start_time = time.time()
# load data
data_root = "pre"
dfTrain = pd.read_csv("%s/train.csv" % data_root)
dfTest = pd.read_csv("%s/test.csv" % data_root)
dfAd = pd.read_csv("%s/ad.csv" % data_root)
dfUser = pd.read_csv("%s/user.csv" % data_root)
dfApp_categories = pd.read_csv("%s/app_categories.csv" % data_root)
# dfPosition = pd.read_csv("%s/position.csv" % data_root)
# dfUser_app_actions = pd.read_csv("%s/user_app_actions.csv" % data_root)
# dfUser_installedapps = pd.read_csv("%s/user_installedapps.csv" % data_root)
"""
# take a look the header to compare
print "Train header: ", dfTrain.head()
print "Test header: ", dfTest.head()
print "Ad header: ", dfAd.head()
print "User header: ", dfUser.head()
print "App_categories header: ", dfApp_categories.head()
print "Postion header: ", dfPosition.head()
print "User_app_actions header: ", dfUser_app_actions.head()
print "User_installedapps header: ", dfUser_installedapps.head()
"""
# process data: join into a big table
dfTrain = pd.merge(dfTrain, dfAd, on="creativeID")
dfTrain = pd.merge(dfTrain, dfUser, on="userID")
dfTrain = pd.merge(dfTrain, dfApp_categories, on="appID")
# dfTrain = pd.merge(dfTrain, dfPosition, on="positionID")
# dfTrain = pd.merge(dfTrain, dfUser_app_actions, on="userID")
# dfTrain = pd.merge(dfTrain, dfUser_installedapps, on="userID")
dfTest = pd.merge(dfTest, dfAd, on="creativeID")
dfTest = pd.merge(dfTest, dfUser, on="userID")
dfTest = pd.merge(dfTest, dfApp_categories, on="appID")
# dfTest = pd.merge(dfTest, dfPosition, on="positionID")
# dfTest = pd.merge(dfTest, dfUser_app_actions, on="userID")
# dfTest = pd.merge(dfTest, dfUser_installedapps, on="userID")
trains = dfTrain # .iloc[:, 0:15]
tests = dfTest # .iloc[:, 0:15]
print '-----------------------------------------------------'
# test_userID = dfTest.userID
# test_creativeID = dfTest.creativeID
params = {
'booster':'gbtree', # 选择每次迭代的模型，有两种选择：1.gbtree：基于树的模型 2.gbliner：线性模型
'objective': 'binary:logistic',
'scale_pos_weight': 1/9.5,
# 27324条正样本
# 218428条负样本
# 差不多1:9/10这样子
'gamma':0.2, # 用于控制是否后剪枝的参数,越大越保守，一般0.1、0.2这样子。
'max_depth':6, # 构建树的深度，越大越容易过拟合
'lambda':3, # 控制模型复杂度的权重值的L2正则化项参数，参数越大，模型越不容易过拟合。
'subsample':0.7, # 随机采样训练样本
#'colsample_bytree':0.7, # 生成树时进行的列采样
'min_child_weight':3,
# 这个参数默认是 1，是每个叶子里面 h 的和至少是多少，对正负样本不均衡时的 0-1 分类而言
#，假设 h 在 0.01 附近，min_child_weight 为 1 意味着叶子节点中最少需要包含 100 个样本。
#这个参数非常影响结果，控制叶子节点中二阶导的和的最小值，该参数值越小，越容易 overfitting。
'silent':0 ,#设置成1则没有运行信息输出，最好是设置为0.
'eta': 0.01, # 如同学习率
'seed':1000,
'nthread':16,# cpu 线程数
'eval_metric': 'rmse'
}
plst = list(params.items())
num_rounds = 1000 # 迭代次数
train_xy, val = train_test_split(trains, test_size=0.15, random_state=1)
# random_state is of big influence for val-auc
y = train_xy.label
X = train_xy.drop(['label', 'conversionTime'], axis=1)
val_y = val.label
val_X = val.drop(['label', 'conversionTime'], axis=1)
test = tests.drop(['instanceID', 'label'], axis=1)
xgb_val = xgb.DMatrix(val_X, label=val_y)
xgb_train = xgb.DMatrix(X, label=y)
xgb_test = xgb.DMatrix(test)
# return 训练和验证的错误率
watchlist = [(xgb_train, 'train'), (xgb_val, 'val')]
print "跑到这里了xgb.train"
# training model
# early_stopping_rounds 当设置的迭代次数较大时，early_stopping_rounds 可在一定的迭代次数内准确率没有提升就停止训练
model = xgb.train(plst, xgb_train, num_rounds, watchlist, early_stopping_rounds=500)
print "跑到这里了save_model"
model.save_model('model/xgb.model') # 用于存储训练出的模型
# print "best best_ntree_limit", model.best_ntree_limit # did not save the best,why?
# print "best best_iteration", model.best_iteration # get it?
print "跑到这里了model.predict"
# preds = model.predict(xgb_test, ntree_limit=model.best_iteration)
# test_y = model.predict(xgb_test, ntree_limit=model.best_iteration)
preds = model.predict(xgb_test, ntree_limit=model.best_iteration)
# labels = xgb_test.get_label()
xgb_result = pd.DataFrame({"instanceID": dfTest["instanceID"].values, "proba": preds})
# xgb_result.proba = preds
xgb_result.sort_values("instanceID", inplace=True)
xgb_result.to_csv("submission.csv", index=False)
with zipfile.ZipFile("submission.zip", "w") as fout:
fout.write("submission.csv", compress_type=zipfile.ZIP_DEFLATED)
# print ('error=%f' % (sum(1 for i in range(len(preds)) if int(preds[i] > 0.5) != labels[i]) / float(len(preds))))
# print ('correct=%f' % (sum(1 for i in range(len(preds)) if int(preds[i] > 0.5) == labels[i]) / float(len(preds))))
"""
test_result = pd.DataFrame(columns=["userID", "creativeID", "label"])
test_result.userID = test_userID
test_result.creativeID = test_creativeID
test_result.label = test_y
test_result.to_csv("output/xgb_bytecup_output.csv", index=None, encoding='utf-8') # remember to edit xgb.csv , add ""
cost_time = time.time() - start_time
print "", '\n', "cost time:", cost_time, "(s)"
# save feature score and feature information: feature,score,min,max,n_null,n_gt1w
feature_score = model.get_fscore()
feature_score = sorted(feature_score.items(), key=lambda x: x[1], reverse=True)
fs = []
for (key, value) in feature_score:
fs.append("{0}, {1}\n".format(key, value))
with open('newfeatures/feature_score_{0}.csv'.format(6), 'w') as f:
f.writelines("feature, score\n")
f.writelines(fs)
# 逻辑回归
# X = X['label_huidalv']
# test = test['label_huidalv']
"""
"""
model1 = LogisticRegression()
model1.fit(X, y)
lr_y1 = model1.predict_log_proba(test)
lr_y2 = model1.predict(test)
lr_y3 = model1.predict_proba(test)
proba_test = model1.predict_proba(test)[:, 1]
lr_result = pd.DataFrame({"instanceID": dfTest["instanceID"].values, "proba": proba_test})
lr_result.sort_values("instanceID", inplace=True)
lr_result.to_csv("submission.csv", index=False)
with zipfile.ZipFile("submission.zip", "w") as fout:
fout.write("submission.csv", compress_type=zipfile.ZIP_DEFLATED)
"""
"""
lr_result.userID = test_userID
lr_result.creativeID = test_creativeID
lr_result.label = lr_y3
lr_result.to_csv("output/lr_bytecup1010_1.csv", index=None, encoding='utf-8') # remember to edit xgb.csv , add ""
"""