合并最近的追溯时间戳并在 pandas 中填充答案

【问题标题】：Merge on nearest retrospective timestamp and fill forward in pandas合并最近的追溯时间戳并在 pandas 中填充
【发布时间】：2021-06-13 07:57:03
【问题描述】：

我很难掌握像 merge_asof() 这样的 pandas 特殊合并功能。

我有两个数据框：coords - 来自 EV gps 的 ping，info - 其他 EV 属性，例如导航目的地和电池电量。我的目标是合并它们，使输出数据帧的行数等于两个数据帧的行数之和。例如：

coords.shape
(10, 3)

coords

ts                          lat       lng
2021-01-02 16:08:24.067971  58.3019 -134.4197
2021-01-06 12:54:18.535681  58.3021 -134.4195
2021-01-08 22:15:35.036423  58.3025 -134.4195
2021-01-16 01:10:39.610540  58.3029 -134.4193
2021-01-27 12:28:45.202376  58.3030 -134.4197
2021-01-30 05:32:09.404525  58.3031 -134.4190
2021-02-08 10:39:19.686159  58.3033 -134.4187
2021-02-15 01:30:16.733921  58.3039 -134.4187
2021-02-16 12:49:55.366025  58.3040 -134.4185
2021-02-19 23:57:57.369978  58.3041 -134.4181


info.shape
(3, 3)

info

ts                          nav_to  battery
2021-01-26 12:47:52.972586  Juneau      90
2021-02-14 23:23:18.186058  Anchorage   50
2021-02-19 07:26:35.357977  Fairbanks   30

info 和 coord 应该合并，以便时间戳 ts 是连续顺序的，并且 info 行应该与 coords 中的行匹配，其中最近的时间戳出现在“之前” .最后，nav_to、battery、lat 和lng 应该向前填写。上述示例的输出将是：

output
ts                          lat      lng        nav_to  battery
2021-01-02 16:08:24.067971  58.3019 -134.4197   None    NaN
2021-01-06 12:54:18.535681  58.3021 -134.4195   None    NaN
2021-01-08 22:15:35.036423  58.3025 -134.4195   None    NaN
2021-01-16 01:10:39.610540  58.3029 -134.4193   None    NaN
2021-01-26 12:47:52.972586  58.3029 -134.4193   Juneau  90.0
2021-01-27 12:28:45.202376  58.3030 -134.4197   Juneau  90.0
2021-01-30 05:32:09.404525  58.3031 -134.4190   Juneau  90.0
2021-02-08 10:39:19.686159  58.3033 -134.4187   Juneau  90.0
2021-02-14 23:23:18.186058  58.3033 -134.4187   Anchorage   50.0
2021-02-15 01:30:16.733921  58.3039 -134.4187   Anchorage   50.0
2021-02-16 12:49:55.366025  58.3040 -134.4185   Anchorage   50.0
2021-02-19 07:26:35.357977  58.3040 -134.4185   Fairbanks   30.0
2021-02-19 23:57:57.369978  58.3041 -134.4181   Fairbanks   30.0

我尝试过使用pd.merge_asof(coords, info, on="ts", direction="forward")，但这不会产生正确的结果，它会向后填充并且只保留来自coords 的记录。在pandas 中产生所需结果的正确命令是什么？

【问题讨论】：

标签： python pandas dataframe merge timestamp

【解决方案1】：

尝试使用默认的direction='backward'，然后使用第二个数据帧concat

(pd.concat([pd.merge_asof(df1, df2, on='ts'), df2])
   .sort_values('ts')
)

输出：

                          ts      lat       lng     nav_to  battery
0 2021-01-02 16:08:24.067971  58.3019 -134.4197        NaN      NaN
1 2021-01-06 12:54:18.535681  58.3021 -134.4195        NaN      NaN
2 2021-01-08 22:15:35.036423  58.3025 -134.4195        NaN      NaN
3 2021-01-16 01:10:39.610540  58.3029 -134.4193        NaN      NaN
0 2021-01-26 12:47:52.972586      NaN       NaN     Juneau     90.0
4 2021-01-27 12:28:45.202376  58.3030 -134.4197     Juneau     90.0
5 2021-01-30 05:32:09.404525  58.3031 -134.4190     Juneau     90.0
6 2021-02-08 10:39:19.686159  58.3033 -134.4187     Juneau     90.0
1 2021-02-14 23:23:18.186058      NaN       NaN  Anchorage     50.0
7 2021-02-15 01:30:16.733921  58.3039 -134.4187  Anchorage     50.0
8 2021-02-16 12:49:55.366025  58.3040 -134.4185  Anchorage     50.0
2 2021-02-19 07:26:35.357977      NaN       NaN  Fairbanks     30.0
9 2021-02-19 23:57:57.369978  58.3041 -134.4181  Fairbanks     30.0

然后您可以选择bfill lat 和lng 列。或者你可以只merge_asof 两次：

(pd.concat([pd.merge_asof(df1, df2, on='ts'), 
            pd.merge_asof(df2, df1, on='ts')
           ])
   .sort_values('ts')
)

输出：

                          ts      lat       lng     nav_to  battery
0 2021-01-02 16:08:24.067971  58.3019 -134.4197        NaN      NaN
1 2021-01-06 12:54:18.535681  58.3021 -134.4195        NaN      NaN
2 2021-01-08 22:15:35.036423  58.3025 -134.4195        NaN      NaN
3 2021-01-16 01:10:39.610540  58.3029 -134.4193        NaN      NaN
0 2021-01-26 12:47:52.972586  58.3029 -134.4193     Juneau     90.0
4 2021-01-27 12:28:45.202376  58.3030 -134.4197     Juneau     90.0
5 2021-01-30 05:32:09.404525  58.3031 -134.4190     Juneau     90.0
6 2021-02-08 10:39:19.686159  58.3033 -134.4187     Juneau     90.0
1 2021-02-14 23:23:18.186058  58.3033 -134.4187  Anchorage     50.0
7 2021-02-15 01:30:16.733921  58.3039 -134.4187  Anchorage     50.0
8 2021-02-16 12:49:55.366025  58.3040 -134.4185  Anchorage     50.0
2 2021-02-19 07:26:35.357977  58.3040 -134.4185  Fairbanks     30.0
9 2021-02-19 23:57:57.369978  58.3041 -134.4181  Fairbanks     30.0

【讨论】：

感谢编辑的答案 - lat/lng NaN 应该如何填写？
@the_darkside se 编辑后的答案:-)。