日期型数据处理

import pandas as pd
#先创建一个数据框（包含缺失值）
df = pd.DataFrame({'auth_date':['2017-01-02','2017-02-02','2017-12-23','NaN'],
                   'sply_date':['2018-01-02','2018-02-02','2018-12-23','NaN'],
                   'rgst_time':['2018-02-03 17:12:42','2018-10-02 12:14:43','2018-03-23 16:23:24','NaN'],
                   'name':['zhangsan','lisi','xiaohua','xiaomei']})

feature = df.columns.tolist()

#当日期型数据比较多时，可以写一个封装好的代码，如下：
def datetime_processing(df):
    """
    argumenr:df:数据框
    goal:       对日期型数据转数值型数据
    return:  df:日期型数据处理完之后的数据
    """
    #日期数据精确到日
    date_feature=['auth_date','sply_date']
    for feature in date_feature:
        df[feature] = pd.to_datetime(df[feature])
        df[feature] = df[feature] - pd.to_datetime("2000-01-01")
        df[feature] = df[feature].astype("str")
        df[feature] = df[feature].apply(lambda x:x.replace("days 00:00:00.000000000","").replace("NaT","0"))
        df[feature] = df[feature].astype("int")
    #日期型数据精确到秒
    datetime_feature = ['rgst_time']
    for feature in datetime_feature:
        df[feature] = pd.to_datetime(df[feature])
        df[feature] = (df[feature] - pd.to_datetime("2000-01-01")).dt.seconds
        df[feature] = df[feature] .fillna(0)
    return df

#看一下处理之后的数据
df = datetime_processing(df)
df.info()

处理前：

df
Out[79]: 
    auth_date   sply_date            rgst_time      name
0  2017-01-02  2018-01-02  2018-02-03 17:12:42  zhangsan
1  2017-02-02  2018-02-02  2018-10-02 12:14:43      lisi
2  2017-12-23  2018-12-23  2018-03-23 16:23:24   xiaohua
3         NaN         NaN                  NaN   xiaomei

处理后：

df
Out[81]: 
   auth_date  sply_date  rgst_time      name
0       6211       6576    61962.0  zhangsan
1       6242       6607    44083.0      lisi
2       6566       6931    59004.0   xiaohua
3          0          0        0.0   xiaomei