在python中计算逐行时间差答案

【问题标题】：Calculating row-wise time difference in python在python中计算逐行时间差
【发布时间】：2020-06-30 12:37:05
【问题描述】：

我想根据他们第一次上公共汽车的时刻和他们离开的时刻之间的差异来计算我的数据框中每个乘客的旅行时间。

这是数据框

my_df = pd.DataFrame({
    'id': ['a', 'b', 'b', 'b', 'b', 'b', 'c','d'],
    'date': ['2020/02/03', '2020/04/05', '2020/04/05', '2020/04/05','2020/04/06', '2020/04/06', '2020/12/15', '2020/06/23'],
    'arriving_time': ['14:36:06', '08:52:02', '08:53:02', '08:55:24', '18:58:03', '19:03:05', '17:04:28', '21:31:23'],
    'leaving_time': ['14:40:05', '08:52:41', '08:54:33', '08:57:14', '19:01:07', '19:04:08', '17:09:48', '21:50:12']
})
print(my_df)

output:

    id  date    arriving_time   leaving_time
0   a   2020/02/03  14:36:06    14:40:05
1   b   2020/04/05  08:52:02    08:52:41
2   b   2020/04/05  08:53:02    08:54:33
3   b   2020/04/05  08:55:24    08:57:14
4   b   2020/04/06  18:58:03    19:01:07
5   b   2020/04/06  19:03:05    19:04:08
6   c   2020/12/15  17:04:28    17:09:48
7   d   2020/06/23  21:31:23    21:50:12

但是有两个问题（我自己无法解决）：

通过手机信号检测到乘客，但信号往往不稳定，这就是为什么对于同一个人，我们可以有很多行（如上述数据集中的乘客 b）。 “arriving_time”是检测到信号的时间，“leaving_time”是信号丢失的时间
要计算行程时间，我需要为每个唯一 ID 和每次行程减去最近的到达时间与最近的离开时间。

这是我想要得到的结果

id  date    arriving_time   leaving_time    travelTime
0   a   2020/02/03  14:36:06    14:40:05    00:03:59
1   b   2020/04/05  08:52:02    08:52:41    00:05:12
2   b   2020/04/05  08:53:02    08:54:33    00:05:12
3   b   2020/04/05  08:55:24    08:57:14    00:05:12
4   b   2020/04/06  18:58:03    19:01:07    00:06:05
5   b   2020/04/06  19:03:05    19:04:08    00:06:05
6   c   2020/12/15  17:04:28    17:09:48    00:05:20
7   d   2020/06/23  21:31:23    21:50:12    00:18:49

如您所见，乘客 b 在同一天进行了两次不同的旅行，我想知道每次旅行的持续时间。

我已经尝试了下面的代码，似乎可以工作，但它确实很慢（我认为这是由于 my_df 的大量行所致）

for user_id in set(my_df.id):
    for day in set(my_df.loc[my_df.id == user_id, 'date']):
        my_df.loc[(my_df.id == user_id) & (my_df.date == day), 'travelTime'] = max(my_df.loc[(my_df.id == user_id) & (my_df.date == day), 'leaving_time'].apply(pd.to_datetime)) - min(my_df.loc[(my_df.id == user_id) & (my_df.date == day), 'arriving_time'].apply(pd.to_datetime))

【问题讨论】：

标签： python pandas numpy datetime time

【解决方案1】：

我认为正确的最大值和最小值是将列转换为日期时间，然后减去由GroupBy.transform 创建的Series：

my_df['s'] = pd.to_datetime(my_df['date'] + ' ' + my_df['arriving_time'])
my_df['e'] = pd.to_datetime(my_df['date'] + ' ' + my_df['leaving_time'])

g = my_df.groupby(['id', 'date'])
my_df['travelTime'] = g['e'].transform('max').sub(g['s'].transform('min'))
print (my_df)
  id        date arriving_time leaving_time                   s  \
0  a  2020/02/03      14:36:06     14:40:05 2020-02-03 14:36:06   
1  b  2020/04/05      08:52:02     08:52:41 2020-04-05 08:52:02   
2  b  2020/04/05      08:53:02     08:54:33 2020-04-05 08:53:02   
3  b  2020/04/05      08:55:24     08:57:14 2020-04-05 08:55:24   
4  b  2020/04/06      18:58:03     19:01:07 2020-04-06 18:58:03   
5  b  2020/04/06      19:03:05     19:04:08 2020-04-06 19:03:05   
6  c  2020/12/15      17:04:28     17:09:48 2020-12-15 17:04:28   
7  d  2020/06/23      21:31:23     21:50:12 2020-06-23 21:31:23   

                    e travelTime  
0 2020-02-03 14:40:05   00:03:59  
1 2020-04-05 08:52:41   00:05:12  
2 2020-04-05 08:54:33   00:05:12  
3 2020-04-05 08:57:14   00:05:12  
4 2020-04-06 19:01:07   00:06:05  
5 2020-04-06 19:04:08   00:06:05  
6 2020-12-15 17:09:48   00:05:20  
7 2020-06-23 21:50:12   00:18:49

为避免出现新列，可以使用DataFrame.assign 系列和datetimes：

s = pd.to_datetime(my_df['date'] + ' ' + my_df['arriving_time'])
e = pd.to_datetime(my_df['date'] + ' ' + my_df['leaving_time'])

g = my_df.assign(s=s, e=e).groupby(['id', 'date'])
my_df['travelTime'] = g['e'].transform('max').sub(g['s'].transform('min'))
print (my_df)
  id        date arriving_time leaving_time travelTime
0  a  2020/02/03      14:36:06     14:40:05   00:03:59
1  b  2020/04/05      08:52:02     08:52:41   00:05:12
2  b  2020/04/05      08:53:02     08:54:33   00:05:12
3  b  2020/04/05      08:55:24     08:57:14   00:05:12
4  b  2020/04/06      18:58:03     19:01:07   00:06:05
5  b  2020/04/06      19:03:05     19:04:08   00:06:05
6  c  2020/12/15      17:04:28     17:09:48   00:05:20
7  d  2020/06/23      21:31:23     21:50:12   00:18:49

【讨论】：

【解决方案2】：

IIUC 我们首先通过groupby id 和date 获取最长和最短离开和到达时间。

然后是一个简单的减法。

df2 = df.groupby(['id','date']).agg(min_arrival=('arriving_time','min'),
                             max_leave=('leaving_time','max'))


df2['travelTime'] =  pd.to_datetime(df2['max_leave']) - pd.to_datetime(df2['min_arrival']) 


print(df2)

              min_arrival max_leave travelTime
id date                                       
a  2020-02-03    14:36:06  14:40:05   00:03:59
b  2020-04-05    08:52:02  08:57:14   00:05:12
   2020-04-06    18:58:03  19:04:08   00:06:05
c  2020-12-15    17:04:28  17:09:48   00:05:20
d  2020-06-23    21:31:23  21:50:12   00:18:49

如果您想在原来的 df 上恢复它，您可以使用 transform 或将新增量中的值合并到原来的：

df_new = (pd.merge(df,df2[['travelTime']],on=['date','id'],how='left')

  id       date arriving_time leaving_time   travelTime
0  a 2020-02-03      14:36:06     14:40:05     00:03:59
1  b 2020-04-05      08:52:02     08:52:41     00:05:12
2  b 2020-04-05      08:53:02     08:54:33     00:05:12
3  b 2020-04-05      08:55:24     08:57:14     00:05:12
4  b 2020-04-06      18:58:03     19:01:07     00:06:05
5  b 2020-04-06      19:03:05     19:04:08     00:06:05
6  c 2020-12-15      17:04:28     17:09:48     00:05:20
7  d 2020-06-23      21:31:23     21:50:12     00:18:49

【讨论】：

【解决方案3】：

你可以试试这个——

my_df['arriving_time'] = pd.to_datetime(my_df['arriving_time'])
my_df['leaving_time'] = pd.to_datetime(my_df['leaving_time'])
my_df['travel_time'] = my_df.groupby(['id', 'date'])['leaving_time'].transform('max') - my_df.groupby(['id', 'date'])['arriving_time'].transform('min')
my_df
    id        date       arriving_time        leaving_time travel_time
0  a  2020/02/03 2020-03-19 14:36:06 2020-03-19 14:40:05    00:03:59
1  b  2020/04/05 2020-03-19 08:52:02 2020-03-19 08:52:41    00:05:12
2  b  2020/04/05 2020-03-19 08:53:02 2020-03-19 08:54:33    00:05:12
3  b  2020/04/05 2020-03-19 08:55:24 2020-03-19 08:57:14    00:05:12
4  b  2020/04/06 2020-03-19 18:58:03 2020-03-19 19:01:07    00:06:05
5  b  2020/04/06 2020-03-19 19:03:05 2020-03-19 19:04:08    00:06:05
6  c  2020/12/15 2020-03-19 17:04:28 2020-03-19 17:09:48    00:05:20
7  d  2020/06/23 2020-03-19 21:31:23 2020-03-19 21:50:12    00:18:49

【讨论】：