我有用户会话ID和发生某些事件的会话ID的时间戳。我想计算第一个事件和最后一个事件之间的时间。请参见下面的示例:
session_id timestamp
sess1 2018-11-05 14:28:25.260
sess2 2018-11-04 12:14:59.576
sess2 2018-11-04 11:55:00.584
sess2 2018-11-04 12:16:44.702
sess3 2018-11-04 12:04:37.419
我想计算sess2的第一个时间戳和最后一个时间戳之间的差,以及所有其他类似的session_id,如下所示:
session_id timeSpent
sess1 1
sess2 125 (for example)
sess3 1
如何计算?
答案 0 :(得分:2)
使用:
#convert column to datetimes if necessary
df['timestamp'] = pd.to_datetime(df['timestamp'])
#aggregate min and max
df1 = df.groupby('session_id')['timestamp'].agg(['min','max'])
#subtract to new column
df1['timeSpent'] = df1.pop('max') - df1.pop('min')
df1 = df1.reset_index()
print (df1)
session_id timeSpent
0 sess1 00:00:00
1 sess2 00:21:44.118000
2 sess3 00:00:00
带有GroupBy.agg
和元组的单行解决方案:
df1 = (df.groupby('session_id')['timestamp']
.agg([('timeSpent', lambda x: x.max() - x.min())])
.reset_index())
print (df1)
session_id timeSpent
0 sess1 00:00:00
1 sess2 00:21:44.118000
2 sess3 00:00:00
如果需要以秒为单位输出,请用Series.dt.total_seconds
转换时间增量:
df1['timeSpent'] = (df1.pop('max') - df1.pop('min')).dt.total_seconds()
df1 = df1.reset_index()
print (df1)
session_id timeSpent
0 sess1 0.000
1 sess2 1304.118
2 sess3 0.000
单行解决方案:
df1 = (df.groupby('session_id')['timestamp']
.agg([('timeSpent', lambda x: x.max() - x.min())])
.assign(timeSpent = lambda x: x['timeSpent'].dt.total_seconds())
.reset_index())
print (df1)
session_id timeSpent
0 sess1 0.000
1 sess2 1304.118
2 sess3 0.000
答案 1 :(得分:1)
您可以将groupby
与apply
结合使用并减去max - min
:
df1 = df.groupby('session_id').timestamp.apply(lambda x: x.max() - x.min()).reset_index()
df1.rename({'timestamp':'timeSpent'},axis=1,inplace=True)
print(df1)
session_id timeSpent
0 sess1 00:00:00
1 sess2 00:21:44.118000
2 sess3 00:00:00
以秒为单位:
df1 = df.groupby('session_id').timestamp.apply(lambda x: x.max() - x.min()).reset_index()
df1.rename({'timestamp':'timeSpent'},axis=1,inplace=True)
df1['timeSpent'] = df1['timeSpent'].dt.total_seconds()
print(df1)
session_id timeSpent
0 sess1 0.000
1 sess2 1304.118
2 sess3 0.000