小花培训笔记

同一个ip过去的登录次数

re=login_table.groupby(['ip'])['event_id'].apply(lambda x:len(np.unique(x))).reset_index()
re1=login_table[['cid','ip']].drop_duplicates()
ip_login_count_data=pd.merge(re1,re,on=['ip'],how='left')

合并数据

login_merge_columns=[pipd,id_count,ip_login_count,country_account,platform_channel,mobile_count,mobilecity_login_account]
login_features=pd.DataFrame(login_table['cid'].unique(),columns=['cid'])
#print(login_features.shape[0])
for i in login_merge_columns:
    login_features=pd.merge(login_features,i,on=['cid'],how='left')
print(login_features.shape[0])

同一个cid每次贷款金额的均值 Loan_mean

def calcute(columns,func,name):
    we=loan_table.groupby(['cid'])[columns].agg(func).reset_index()
    we1=we.rename(columns={columns:name})
    return we1
Loan_mean=calcute('loanAmount','mean','Loan_mean')
	cid						Loan_mean
0	20170801000012728075	9468.000
1	20170801000012728093	8876.250

计算KS值的标准代码

from scipy.stats import ks_2samp
get_ks = lambda y_pred,y_true: ks_2samp(y_pred[y_true==1], y_pred[y_true!=1]).statistic
get_ks(x,y)