section 7

最新推荐文章于 2020-11-28 03:26:31 发布

weixin_30817749

最新推荐文章于 2020-11-28 03:26:31 发布

阅读量241

点赞数

文章标签：人工智能数据结构与算法数据库

原文链接：http://www.cnblogs.com/cvlas/p/9529061.html

版权

import pandas as pd
import numpy as np
import pymysql
from sqlalchemy import create_engine
import matplotlib.pyplot as plt
import seaborn as sns
import missingno as msno
%matplotlib inline

# 数据库引擎
engine = create_engine('mysql+pymysql://root:123456@localhost:3306/datascience')

读取数据

# 读取数据
data = 'data/section7-dau.csv'

dau = pd.read_csv(data)

# 写入 MYSQL
# dau.to_sql('s7_dau',engine,index=False)

dau.head()

	region_month	region_day	app_name	user_id	device
0	2013-01	2013-01-01	game-02	10061580	FP
1	2013-01	2013-01-01	game-02	10154440	FP
2	2013-01	2013-01-01	game-02	10164762	SP
3	2013-01	2013-01-01	game-02	10165615	FP
4	2013-01	2013-01-01	game-02	10321356	FP

# 查看数据信息
dau.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 48988 entries, 0 to 48987
Data columns (total 5 columns):
region_month    48988 non-null object
region_day      48988 non-null object
app_name        48988 non-null object
user_id         48988 non-null int64
device          48988 non-null object
dtypes: int64(1), object(4)
memory usage: 1.9+ MB

print(dau.region_month.value_counts())
print(dau.region_day.unique())
print(dau.device.value_counts())

2013-01    25847
2013-02    23141
Name: region_month, dtype: int64
['2013-01-01' '2013-01-02' '2013-01-03' '2013-01-04' '2013-01-05'
 '2013-01-06' '2013-01-07' '2013-01-08' '2013-01-09' '2013-01-10'
 '2013-01-11' '2013-01-12' '2013-01-13' '2013-01-14' '2013-01-15'
 '2013-01-16' '2013-01-17' '2013-01-18' '2013-01-19' '2013-01-20'
 '2013-01-21' '2013-01-22' '2013-01-23' '2013-01-24' '2013-01-25'
 '2013-01-26' '2013-01-27' '2013-01-28' '2013-01-29' '2013-01-30'
 '2013-01-31' '2013-02-01' '2013-02-02' '2013-02-03' '2013-02-04'
 '2013-02-05' '2013-02-06' '2013-02-07' '2013-02-08' '2013-02-09'
 '2013-02-10' '2013-02-11' '2013-02-12' '2013-02-13' '2013-02-14'
 '2013-02-15' '2013-02-16' '2013-02-17' '2013-02-18' '2013-02-19'
 '2013-02-20' '2013-02-21' '2013-02-22' '2013-02-23' '2013-02-24'
 '2013-02-25' '2013-02-26' '2013-02-27' '2013-02-28']
FP    30331
SP    18657
Name: device, dtype: int64

关于用户是否进行了账号迁转的数据的整理

提取需要的数据列，去除重复项，得到用户按月份和设备登陆的信息

mau = dau[['region_month','user_id','device']]
mau.head()

	region_month	user_id	device
0	2013-01	10061580	FP
1	2013-01	10154440	FP
2	2013-01	10164762	SP
3	2013-01	10165615	FP
4	2013-01	10321356	FP

# 重复数据 （用户在某一月使用相同设备登陆）
print(mau.duplicated().sum())
mau.drop_duplicates(inplace=True)
print(mau.duplicated().sum())

46007
0


D:\ProgramData\Anaconda3\lib\site-packages\ipykernel_launcher.py:3: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
  This is separate from the ipykernel package so we can avoid doing imports until

非智能手机和智能手机分开

fp = dau[dau['device']=='FP'][['region_month','user_id','device']].drop_duplicates()
sp = dau[dau['device']=='SP'][['region_month','user_id','device']].drop_duplicates()

print(fp.info())
print(sp.info())

<class 'pandas.core.frame.DataFrame'>
Int64Index: 1755 entries, 0 to 48901
Data columns (total 3 columns):
region_month    1755 non-null object
user_id         1755 non-null int64
device          1755 non-null object
dtypes: int64(1), object(2)
memory usage: 54.8+ KB
None
<class 'pandas.core.frame.DataFrame'>
Int64Index: 1226 entries, 2 to 48834
Data columns (total 3 columns):
region_month    1226 non-null object
user_id         1226 non-null int64
device          1226 non-null object
dtypes: int64(1), object(2)
memory usage: 38.3+ KB
None

分别获取1月份和2月份的数据

# 分别获取1月份和2月份的数据

fp_m1 = fp[fp['region_month']=='2013-01']
fp_m2 = fp[fp['region_month']=='2013-02']

sp_m1 = sp[sp['region_month']=='2013-01']
sp_m2 = sp[sp['region_month']=='2013-02']

1月份的非智能手机用户在2月份的访问情况

# 1月份的非智能手机用户在2月份的访问情况

mau['is_access'] = 1
fp_m1 = pd.merge(fp_m1,mau[mau['region_month']=='2013-02'][['user_id','is_access']],how='left',on='user_id')
fp_m1['is_access'].fillna(0,inplace=True)

fp_m1.head()

D:\ProgramData\Anaconda3\lib\site-packages\ipykernel_launcher.py:3: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
  This is separate from the ipykernel package so we can avoid doing imports until

	region_month	user_id	device	is_access
0	2013-01	10061580	FP	1.0
1	2013-01	10154440	FP	0.0
2	2013-01	10165615	FP	1.0
3	2013-01	10321356	FP	1.0
4	2013-01	10447112	FP	1.0

1月份访问过游戏的非智能手机用户在2月份是否是继续通过非智能手机来访问的

# 1月份访问过游戏的非智能手机用户在2月份是否是继续通过非智能手机来访问的

fp_m2['is_fp'] = 1
fp_m1 = pd.merge(fp_m1,fp_m2[['user_id','is_fp']],how='left',on='user_id')
fp_m1['is_fp'].fillna(0,inplace=True)

fp_m1.head()

D:\ProgramData\Anaconda3\lib\site-packages\ipykernel_launcher.py:3: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
  This is separate from the ipykernel package so we can avoid doing imports until

	region_month	user_id	device	is_access	is_fp
0	2013-01	10061580	FP	1.0	1.0
1	2013-01	10154440	FP	0.0	0.0
2	2013-01	10165615	FP	1.0	1.0
3	2013-01	10321356	FP	1.0	1.0
4	2013-01	10447112	FP	1.0	1.0

1月份访问过游戏的非智能手机用户在2月份是否是通过智能手机来访问的

# 1月份访问过游戏的非智能手机用户在2月份是否是通过智能手机来访问的

sp_m2['is_sp'] = 1
fp_m1 = pd.merge(fp_m1,sp_m2[['user_id','is_sp']],how='left',on='user_id')
fp_m1['is_sp'].fillna(0,inplace=True)

fp_m1.head()

D:\ProgramData\Anaconda3\lib\site-packages\ipykernel_launcher.py:3: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
  This is separate from the ipykernel package so we can avoid doing imports until

	region_month	user_id	device	is_access	is_fp
0	2013-01	10061580	FP	1.0	1.0
1	2013-01	10154440	FP	0.0	0.0
2	2013-01	10165615	FP	1.0	1.0
3	2013-01	10321356	FP	1.0	1.0
4	2013-01	10447112	FP	1.0	1.0

1月份通过非智能手机访问但2月份没有访问的用户，或者通过智能手机访问的用户

# 1月份通过非智能手机访问但2月份没有访问的用户，或者通过智能手机访问的用户

fp_m1 = fp_m1[(fp_m1['is_access']==0) | (fp_m1['is_sp']==1)]
fp_m1.head()

	region_month	user_id	device	is_access	is_sp
1	2013-01	10154440	FP	0.0	0.0
7	2013-01	10528830	FP	0.0	0.0
20	2013-01	1163733	FP	1.0	1.0
21	2013-01	11727630	FP	0.0	0.0
43	2013-01	13401362	FP	1.0	1.0

以上得到的即是可用于逻辑回归的标签项

关于是否是每天访问游戏的数据的整理

# 标记每天登陆记录

fp_dau = dau[(dau['device']=='FP') & (dau['region_month']=='2013-01')]
fp_dau['is_access'] = 1
fp_dau.head()

D:\ProgramData\Anaconda3\lib\site-packages\ipykernel_launcher.py:4: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
  after removing the cwd from sys.path.

	region_month	region_day	app_name	user_id	device	is_access
0	2013-01	2013-01-01	game-02	10061580	FP	1
1	2013-01	2013-01-01	game-02	10154440	FP	1
3	2013-01	2013-01-01	game-02	10165615	FP	1
4	2013-01	2013-01-01	game-02	10321356	FP	1
6	2013-01	2013-01-01	game-02	10447112	FP	1

# COLUMNS 名字
b = []
for a in np.arange(1,32):
    b.append('X'+str(a)+'day')
# b.insert(0,'user_id')

# 透视表转化为登陆信息
fp_dau_pivot = pd.pivot_table(fp_dau, values='is_access', columns='region_day', index='user_id', fill_value=0)
fp_dau_pivot.columns = b
fp_dau_pivot.reset_index(inplace=True)

fp_dau_pivot.head()

	user_id	X1day	X2day	X3day	X4day	X5day	X6day	X7day	X8day	X9day	...	X22day	X23day	X24day	X25day	X26day	X27day	X28day	X29day	X30day	X31day
0	397286	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1	1	1
1	471341	1	1	1	1	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
2	503874	1	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
3	512250	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1	1	1
4	513811	0	0	0	0	0	0	0	0	0	...	1	0	0	0	0	0	1	1	0	1

5 rows × 32 columns

# 将2月份的访问数据和智能手机用户数据合并,注意这里是 inner 。。

fp_dau_m = pd.merge(fp_dau_pivot, fp_m1[['user_id','is_sp']], how='inner', on='user_id')

fp_dau_m.head()

	user_id	X1day	X2day	X3day	X4day	...	X23day	X24day	X25day	X26day	X27day	X28day	is_sp
0	471341	1	1	1	1	...	0	0	0	0	0	0	1.0
1	503874	1	0	0	0	...	0	0	0	0	0	0	0.0
2	1073544	0	0	0	0	...	1	1	1	0	0	0	0.0
3	1073864	0	0	0	0	...	0	0	0	0	0	0	0.0
4	1163733	1	1	0	0	...	1	1	1	1	1	1	1.0

5 rows × 33 columns

fp_dau_m.isna().sum().sum()

fp_dau_m.is_sp.value_counts()

0.0    190
1.0     62
Name: is_sp, dtype: int64

以上数据显示，is_sp 指示： 1表示2月份通过智能手机来访问的用户， 0表示用户为流失用户

2月份流失的用户数有190个，更换为智能手机用户数为62个！

逻辑回归处理

1.sklearn

通过修改 solve 和惩罚系数 C ，可以将模型的准确度提升至 100%

from sklearn.linear_model import LogisticRegression

lr = LogisticRegression(solver='lbfgs',C=10)

x = fp_dau_m.iloc[:,1:-1]
y = fp_dau_m.iloc[:,-1]

lr.fit(x,y)

print('系数项：',lr.coef_)
print('截距项：',lr.intercept_)
print('得分是：',lr.score(x,y))

系数项： [[ 1.64264315  0.38232509  0.27375659  1.77818234 -1.2604587  -0.62425027
   1.64964331  0.94366796 -0.30971957 -2.45689215  1.05453162 -0.49567095
   1.37452985 -0.79198757 -1.39648934  0.18038175 -0.34026571  1.01401641
  -0.49919155 -0.25791649  0.98296119  1.03952236 -1.03446927  1.53177282
  -0.12212919  0.30942289  0.31267693 -0.08203749  1.32893163  1.57890364
   1.29380472]]
截距项： [-3.9031072]
得分是： 0.9047619047619048

yp = lr.predict_proba(x)[:,1]

df = fp_dau_m.copy()
df['prob'] = yp
df['pred'] = df['prob'].apply(lambda x: 1 if x > 0.5 else 0)
df.head(15)

	user_id	X1day	X2day	X3day	X4day	X5day	X6day	X8day	X9day	...	X25day	X26day	X27day	X28day	X29day	X30day	X31day	is_sp	prob	pred
0	471341	1	1	1	1	0	0	0	0	...	0	0	0	0	0	0	0	1.0	0.543341	1
1	503874	1	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.094451	0
2	1073544	0	0	0	0	0	0	0	0	...	1	0	0	0	0	0	0	0.0	0.002510	0
3	1073864	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.025567	0
4	1163733	1	1	0	0	0	0	0	0	...	1	1	1	1	0	0	0	1.0	0.849838	1
5	1454629	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.073879	0
6	1557628	0	0	0	0	0	0	0	1	...	0	0	0	0	0	0	1	0.0	0.051221	0
7	2241462	1	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	1.0	0.094451	0
8	2313236	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.085385	0
9	2477685	0	0	0	0	0	0	0	0	...	1	0	0	0	0	0	0	0.0	0.017546	0
10	2541741	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.001726	0
11	2628661	0	0	0	0	0	1	0	0	...	0	1	0	0	0	0	0	0.0	0.014515	0
12	3509436	0	1	0	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.987940	1
13	3509436	0	1	0	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.987940	1
14	3955950	1	1	1	1	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.543341	1

15 rows × 35 columns

df.groupby(['is_sp','pred'])['user_id'].count().reset_index()

	is_sp	pred	user_id
0	0.0	0	181
1	0.0	1	9
2	1.0	0	15
3	1.0	1	47

len(df[df['is_sp']==df['pred']])/len(df)

0.9047619047619048

此模型，无需修改任何参数即可达到准确度 100% 。重点在于 solve 和 C 的参数。

from sklearn.linear_model import LogisticRegressionCV

lr = LogisticRegressionCV(cv=10)

x = fp_dau_m.iloc[:,1:-1]
y = fp_dau_m.iloc[:,-1]

lr.fit(x,y)

print('系数项：',lr.coef_)
print('截距项：',lr.intercept_)
print('-----------------------------------------------')
print('得分是: ',lr.score(x,y))

系数项： [[ 0.66247469  0.39566209  0.12089587  0.72621501 -0.14485039 -0.11496137
   0.50433275  0.25667173  0.11561233 -0.48159577  0.23713178 -0.12897139
   0.31542595 -0.16714406 -0.1914315  -0.09390318 -0.05036135  0.0924934
  -0.14949742 -0.05918408  0.52355482  0.58543392  0.0882812   0.39783666
   0.07477356  0.14874974  0.39921228  0.38402639  0.68729765  0.6331324
   0.55885631]]
截距项： [-2.95546571]
-----------------------------------------------
得分是:  0.8928571428571429

statsmodels

import statsmodels.api as sm
import statsmodels.formula.api as fsm     # 这个是用公式去拟合，不好用

x = fp_dau_m.iloc[:,1:-1]
x['intercept'] = 1.0          # 此处是为logistics回归添加截距项
y = fp_dau_m.iloc[:,-1]

logit = sm.Logit(y, x)
result = logit.fit(method='bfgs',maxiter=100)

Warning: Maximum number of iterations has been exceeded.
         Current function value: 0.222887
         Iterations: 100
         Function evaluations: 101
         Gradient evaluations: 101


C:\Users\sylva\AppData\Roaming\Python\Python36\site-packages\statsmodels\base\model.py:508: ConvergenceWarning: Maximum Likelihood optimization failed to converge. Check mle_retvals
  "Check mle_retvals", ConvergenceWarning)

# result1 = logit.fit_regularized(alpha=5)

result.pred_table()

array([[180.,  10.],
       [ 14.,  48.]])

# result1.pred_table()

print(result.summary2())

                         Results: Logit
=================================================================
Model:              Logit            Pseudo R-squared: 0.601     
Dependent Variable: is_sp            AIC:              176.3352  
Date:               2018-08-24 12:07 BIC:              289.2770  
No. Observations:   252              Log-Likelihood:   -56.168   
Df Model:           31               LL-Null:          -140.60   
Df Residuals:       220              LLR p-value:      6.6358e-21
Converged:          0.0000           Scale:            1.0000    
------------------------------------------------------------------
               Coef.   Std.Err.     z     P>|z|    [0.025   0.975]
------------------------------------------------------------------
X1day          1.9894    0.8047   2.4720  0.0134   0.4121   3.5666
X2day          0.3311    1.0705   0.3093  0.7571  -1.7671   2.4293
X3day          0.3793    0.9406   0.4033  0.6867  -1.4641   2.2227
X4day          2.0422    0.8359   2.4430  0.0146   0.4038   3.6805
X5day         -1.7597    1.1991  -1.4675  0.1422  -4.1100   0.5906
X6day         -0.6679    1.1717  -0.5701  0.5686  -2.9643   1.6285
X7day          2.0157    1.1176   1.8036  0.0713  -0.1747   4.2061
X8day          1.2119    1.3505   0.8974  0.3695  -1.4350   3.8589
X9day         -0.4495    1.1874  -0.3786  0.7050  -2.7768   1.8778
X10day        -3.2374    1.5580  -2.0779  0.0377  -6.2911  -0.1837
X11day         1.4392    1.2234   1.1764  0.2394  -0.9586   3.8370
X12day        -0.6389    1.5297  -0.4176  0.6762  -3.6370   2.3592
X13day         1.7797    1.1424   1.5579  0.1193  -0.4594   4.0188
X14day        -1.1242    1.2455  -0.9026  0.3668  -3.5653   1.3170
X15day        -1.8115    1.3050  -1.3881  0.1651  -4.3694   0.7463
X16day         0.4940    1.1666   0.4234  0.6720  -1.7925   2.7804
X17day        -0.4448    1.2234  -0.3636  0.7162  -2.8427   1.9531
X18day         1.4321    1.1465   1.2491  0.2116  -0.8150   3.6791
X19day        -0.6132    1.1990  -0.5114  0.6091  -2.9632   1.7369
X20day        -0.3130    1.4007  -0.2235  0.8232  -3.0585   2.4324
X21day         0.9587    1.2558   0.7634  0.4452  -1.5027   3.4201
X22day         1.1954    1.1238   1.0637  0.2875  -1.0072   3.3980
X23day        -1.5371    1.2303  -1.2494  0.2115  -3.9486   0.8743
X24day         1.8445    1.1038   1.6710  0.0947  -0.3190   4.0080
X25day         0.1292    1.5317   0.0844  0.9328  -2.8727   3.1312
X26day         0.3131    1.4280   0.2192  0.8265  -2.4858   3.1119
X27day         0.3365    1.2965   0.2596  0.7952  -2.2045   2.8776
X28day        -0.3918    1.8515  -0.2116  0.8324  -4.0207   3.2372
X29day         1.5941    1.0565   1.5088  0.1314  -0.4767   3.6648
X30day         1.9943    1.2117   1.6459  0.0998  -0.3806   4.3692
X31day         1.5214    1.1798   1.2896  0.1972  -0.7908   3.8337
intercept     -4.2502    0.5904  -7.1985  0.0000  -5.4074  -3.0930
=================================================================

# print(result1.summary2())

xx = fp_dau_m.iloc[:,1:-1]
xx['intercept'] = 1.0         # 预测也要为logistics回归添加截距项

y_p = result.predict(xx)

ydf = fp_dau_m.copy()
ydf['prob'] = y_p
ydf['pred'] = ydf['prob'].apply(lambda x: 1 if x > 0.5 else 0)
ydf.head(15)

	user_id	X1day	X2day	X3day	X4day	X5day	X6day	X8day	X9day	...	X25day	X26day	X27day	X28day	X29day	X30day	X31day	is_sp	prob	pred
0	471341	1	1	1	1	0	0	0	0	...	0	0	0	0	0	0	0	1.0	0.620506	1
1	503874	1	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.094416	0
2	1073544	0	0	0	0	0	0	0	0	...	1	0	0	0	0	0	0	0.0	0.000866	0
3	1073864	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.019167	0
4	1163733	1	1	0	0	0	0	0	0	...	1	1	1	1	0	0	0	1.0	0.870576	1
5	1454629	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.077951	0
6	1557628	0	0	0	0	0	0	0	1	...	0	0	0	0	0	0	1	0.0	0.039991	0
7	2241462	1	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	1.0	0.094416	0
8	2313236	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.082739	0
9	2477685	0	0	0	0	0	0	0	0	...	1	0	0	0	0	0	0	0.0	0.015969	0
10	2541741	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.000560	0
11	2628661	0	0	0	0	0	1	0	0	...	0	1	0	0	0	0	0	0.0	0.009902	0
12	3509436	0	1	0	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.992456	1
13	3509436	0	1	0	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.992456	1
14	3955950	1	1	1	1	0	0	0	0	...	0	0	0	0	0	0	0	0.0	0.620506	1

15 rows × 35 columns

ydf.groupby(['is_sp','pred'])['user_id'].count().reset_index()

	is_sp	pred	user_id
0	0.0	0	180
1	0.0	1	10
2	1.0	0	14
3	1.0	1	48

len(ydf[ydf['is_sp']==ydf['pred']])/len(ydf)

0.9047619047619048

结果观察

根据 sklearn 预测的结果，有9名用户预测为1，即进行了账号迁转，但实际并没有。根据过去的访问情况来推断，这些用户应该进行了账号迁转，然而实际却是流失的用户群体。

df.head(10)

	user_id	X1day	X2day	X3day	X4day	X9day	...	X25day	X26day	X27day	X28day	X31day	is_sp	prob	pred
0	471341	1	1	1	1	0	...	0	0	0	0	0	1.0	0.543341	1
1	503874	1	0	0	0	0	...	0	0	0	0	0	0.0	0.094451	0
2	1073544	0	0	0	0	0	...	1	0	0	0	0	0.0	0.002510	0
3	1073864	0	0	0	0	0	...	0	0	0	0	0	0.0	0.025567	0
4	1163733	1	1	0	0	0	...	1	1	1	1	0	1.0	0.849838	1
5	1454629	0	0	0	0	0	...	0	0	0	0	0	0.0	0.073879	0
6	1557628	0	0	0	0	1	...	0	0	0	0	1	0.0	0.051221	0
7	2241462	1	0	0	0	0	...	0	0	0	0	0	1.0	0.094451	0
8	2313236	0	0	0	0	0	...	0	0	0	0	0	0.0	0.085385	0
9	2477685	0	0	0	0	0	...	1	0	0	0	0	0.0	0.017546	0

10 rows × 35 columns

df1 = df[(df['is_sp']==1) & (df['pred']==1)]
df1.sort_values(by='prob',ascending=True).head(15)

	user_id	X1day	X2day	X3day	X4day	X5day	X6day	X7day	X8day	X9day	...	X25day	X26day	X27day	X28day	X29day	X30day	X31day	is_sp	prob	pred
228	52776438	1	1	1	1	1	1	1	1	1	...	0	0	0	0	0	0	0	1.0	0.512293	1
171	32762652	1	1	1	1	1	1	1	1	1	...	0	0	0	0	0	0	0	1.0	0.512293	1
155	27800629	1	1	1	1	0	0	0	0	0	...	0	0	0	0	0	0	0	1.0	0.543341	1
0	471341	1	1	1	1	0	0	0	0	0	...	0	0	0	0	0	0	0	1.0	0.543341	1
36	8645980	0	0	0	1	0	0	0	0	0	...	0	0	0	0	1	0	0	1.0	0.551574	1
37	8645980	0	0	0	1	0	0	0	0	0	...	0	0	0	0	1	0	0	1.0	0.551574	1
169	32500332	1	1	1	1	1	1	1	1	1	...	1	0	0	0	0	0	0	1.0	0.587923	1
55	11600349	0	1	1	1	1	1	1	1	1	...	0	0	0	1	1	1	1	1.0	0.684198	1
56	11600349	0	1	1	1	1	1	1	1	1	...	0	0	0	1	1	1	1	1.0	0.684198	1
146	25787360	0	0	0	0	1	0	1	1	1	...	0	0	1	0	0	0	0	1.0	0.696295	1
145	25787360	0	0	0	0	1	0	1	1	1	...	0	0	1	0	0	0	0	1.0	0.696295	1
4	1163733	1	1	0	0	0	0	0	0	0	...	1	1	1	1	0	0	0	1.0	0.849838	1
48	10406653	0	1	1	1	1	1	1	1	0	...	1	0	1	1	1	1	1	1.0	0.865393	1
49	10406653	0	1	1	1	1	1	1	1	0	...	1	0	1	1	1	1	1	1.0	0.865393	1
165	31066299	0	1	1	1	0	1	1	1	1	...	1	1	1	0	1	1	0	1.0	0.951970	1

15 rows × 35 columns

df2 = df[(df['is_sp']==1) & (df['pred']==1)]
df2.sort_values(by='prob',ascending=False).head(15)

	user_id	X1day	X2day	X3day	X4day	X5day	X6day	X7day	X8day	X9day	...	X25day	X26day	X27day	X28day	X29day	X30day	X31day	is_sp	prob	pred
136	24791702	1	1	0	1	0	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.998618	1
137	24791702	1	1	0	1	0	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.998618	1
44	9567562	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.996302	1
43	9567562	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.996302	1
139	24900784	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1
124	23113079	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1
133	24581383	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1
134	24581383	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1
138	24900784	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1
123	23113079	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1
114	21551429	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1
147	27003770	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1
148	27003770	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1
150	27602710	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1
151	27602710	1	1	1	1	1	1	1	1	1	...	1	1	1	1	1	1	1	1.0	0.993923	1

15 rows × 35 columns

df3 = df[(df['is_sp']==0) & (df['pred']==1)]
df3.sort_values(by='prob',ascending=False).head(15)

	user_id	X1day	X2day	X3day	X4day	X5day	X6day	X7day	X8day	X9day	...	X25day	X26day	X27day	X28day	X29day	X31day	prob	pred
194	41590801	0	0	0	0	0	0	0	0	0	...	0	0	0	0	1	1	0.677458	1
108	19432099	1	1	1	1	0	1	1	1	1	...	0	0	0	0	0	0	0.643061	1
203	43451947	1	1	1	1	1	0	1	1	1	...	1	0	0	1	1	0	0.599921	1
197	42276142	1	1	1	1	1	1	0	1	1	...	1	1	1	1	1	0	0.577420	1
209	46285446	0	0	0	0	1	1	1	1	1	...	1	1	1	0	1	0	0.576873	1
14	3955950	1	1	1	1	0	0	0	0	0	...	0	0	0	0	0	0	0.543341	1
158	28391896	1	1	1	1	1	1	1	1	1	...	0	0	0	0	0	0	0.512293	1
240	59561276	1	1	1	1	1	1	1	1	1	...	0	0	0	0	0	0	0.512293	1
27	6147878	1	0	0	1	1	1	1	1	1	...	1	1	0	0	0	0	0.502182	1

9 rows × 35 columns

df4 = df[(df['is_sp']==0) & (df['pred']==1)]
df4.sort_values(by='prob',ascending=True).head(15)

	user_id	X1day	X2day	X3day	X4day	X5day	X6day	X7day	X8day	X9day	...	X25day	X26day	X27day	X28day	X29day	X31day	prob	pred
27	6147878	1	0	0	1	1	1	1	1	1	...	1	1	0	0	0	0	0.502182	1
158	28391896	1	1	1	1	1	1	1	1	1	...	0	0	0	0	0	0	0.512293	1
240	59561276	1	1	1	1	1	1	1	1	1	...	0	0	0	0	0	0	0.512293	1
14	3955950	1	1	1	1	0	0	0	0	0	...	0	0	0	0	0	0	0.543341	1
209	46285446	0	0	0	0	1	1	1	1	1	...	1	1	1	0	1	0	0.576873	1
197	42276142	1	1	1	1	1	1	0	1	1	...	1	1	1	1	1	0	0.577420	1
203	43451947	1	1	1	1	1	0	1	1	1	...	1	0	0	1	1	0	0.599921	1
108	19432099	1	1	1	1	0	1	1	1	1	...	0	0	0	0	0	0	0.643061	1
194	41590801	0	0	0	0	0	0	0	0	0	...	0	0	0	0	1	1	0.677458	1

9 rows × 35 columns

df5 = df[(df['is_sp']==0) & (df['pred']==0)]
df5.sort_values(by='prob',ascending=True).head(15)

	user_id	X4day	X5day	X6day	X7day	...	X25day	prob
149	27249550	1	1	1	0	...	0	0.000946
10	2541741	0	0	0	0	...	0	0.001726
242	60725457	0	0	0	0	...	0	0.001726
101	18408297	0	0	0	0	...	0	0.001745
172	33766090	0	0	0	0	...	0	0.002257
2	1073544	0	0	0	0	...	1	0.002510
227	52612953	0	0	0	0	...	0	0.003087
63	12582684	1	1	0	1	...	0	0.004780
208	46056688	0	0	1	1	...	0	0.004799
66	13157777	0	0	0	0	...	0	0.004969
190	40654033	0	0	0	0	...	0	0.004969
120	22437652	0	1	0	0	...	0	0.005689
87	16601600	0	1	0	0	...	0	0.005689
70	13967453	0	1	0	0	...	0	0.005689
112	20955934	0	1	0	0	...	0	0.005689

15 rows × 35 columns

df6 = df[(df['is_sp']==1) & (df['pred']==0)]
df6.sort_values(by='prob',ascending=False).head(15)

	user_id	X1day	X2day	X3day	X4day	X5day	X6day	X7day	X8day	X9day	...	X25day	X26day	X27day	X29day	is_sp	prob
198	42438713	1	1	1	1	1	1	1	0	0	...	0	0	0	0	1.0	0.484688
127	23689923	1	1	0	1	1	1	1	1	1	...	0	0	0	0	1.0	0.359100
213	47332069	0	0	0	0	0	0	0	0	0	...	1	1	0	0	1.0	0.281079
140	24914421	1	1	1	0	0	0	0	1	0	...	0	1	0	0	1.0	0.278119
226	52131958	0	0	1	1	1	1	1	1	1	...	1	1	1	0	1.0	0.259709
212	47266966	1	0	0	1	0	1	1	1	1	...	0	0	0	0	1.0	0.232730
236	57869405	0	0	0	0	0	0	1	1	0	...	0	0	0	0	1.0	0.212521
161	29698758	1	1	1	0	0	0	0	0	0	...	0	0	0	0	1.0	0.167370
30	7177251	1	1	1	1	1	1	0	0	0	...	0	0	0	0	1.0	0.153046
7	2241462	1	0	0	0	0	0	0	0	0	...	0	0	0	0	1.0	0.094451
67	13401362	1	0	0	0	0	0	0	0	0	...	0	0	0	0	1.0	0.094451
80	15569351	0	0	0	0	0	0	1	0	1	...	0	0	0	0	1.0	0.071546
93	17388480	0	0	0	0	0	0	0	0	0	...	0	0	0	1	1.0	0.070819
94	17388480	0	0	0	0	0	0	0	0	0	...	0	0	0	1	1.0	0.070819
163	30103279	0	0	0	0	0	0	0	0	0	...	0	0	0	0	1.0	0.028795

15 rows × 35 columns

copy 问题的出现了，！！！ = 等号只是引用内存地址，变量最好用 copy() 属性！！

fp_dau_m.head()

	user_id	X1day	X2day	X3day	X4day	...	X23day	X24day	X25day	X26day	X27day	X28day	is_sp
0	471341	1	1	1	1	...	0	0	0	0	0	0	1.0
1	503874	1	0	0	0	...	0	0	0	0	0	0	0.0
2	1073544	0	0	0	0	...	1	1	1	0	0	0	0.0
3	1073864	0	0	0	0	...	0	0	0	0	0	0	0.0
4	1163733	1	1	0	0	...	1	1	1	1	1	1	1.0

5 rows × 33 columns

df.equals(fp_dau_m)

False

df.equals(ydf)

False

转载于:https://www.cnblogs.com/cvlas/p/9529061.html

weixin_30817749

关注

0
点赞
踩
0

收藏

觉得还不错? 一键收藏
0
评论
section 7

import pandas as pdimport numpy as npimport pymysqlfrom sqlalchemy import create_engineimport matplotlib.pyplot as pltimport seaborn as snsimport missingno as msno%matplotlib inline# 数据库引...
复制链接

扫一扫