关于 PYTHON 中 2 个字段的每行之间的持续时间总和

Question

我有一组关于时间戳、模型和模式的数据 该模式有 4 种不同的模式，表示为 (0,2,4,8)

Index  Model Timestamp              Mode
1      x  2016-06-26 09:51:24.279   0
2      x  2016-06-26 09:51:26.282   0
3      x  2016-06-26 09:51:28.279   0
4      x  2016-06-26 09:51:30.279   0
5      y  2016-06-26 09:51:32.279   2
6      y  2016-06-26 09:51:34.279   2
7      x  2016-06-26 09:51:36.278   2
8      x  2016-06-26 09:51:38.279   0
9      x  2016-06-26 09:51:40.279   0
10     y  2016-06-26 09:51:42.280   0
11     x  2016-06-26 09:51:44.279   2
12     x  2016-06-26 09:51:46.279   0
13     x  2016-06-26 09:51:48.279   2
14     x  2016-06-26 09:51:50.281   0
15     y  2016-06-26 09:51:52.279   2
16     y  2016-06-26 09:51:54.279   2
17     x  2016-06-26 09:51:56.281   0
18     x  2016-06-26 09:51:58.279   0
19     x  2016-06-26 09:52:00.279   0

我会想在他们的模式（映射到模型）中寻找时差。因此，如果它从模式 0 过渡到模式 2，则在第一个模式 2 应该计算 0，它应该计算持续时间差异，如下所示

Index  Model Timestamp              Mode  Diff
1      x  2016-06-26 09:51:24.279   0     0
2      x  2016-06-26 09:51:26.282   0     2.002
3      x  2016-06-26 09:51:28.279   0     1.998
4      x  2016-06-26 09:51:30.279   0     2
5      y  2016-06-26 09:51:32.279   2     0
6      y  2016-06-26 09:51:34.279   2     2
7      x  2016-06-26 09:51:36.278   2     0
8      x  2016-06-26 09:51:38.279   0     0
9      x  2016-06-26 09:51:40.279   0     2
10     y  2016-06-26 09:51:42.280   0     0
11     x  2016-06-26 09:51:44.279   2     0
12     x  2016-06-26 09:51:46.279   0     0
13     x  2016-06-26 09:51:48.279   2     0
14     x  2016-06-26 09:51:50.281   0     0
15     y  2016-06-26 09:51:52.279   2     0
16     y  2016-06-26 09:51:54.279   2     2
17     x  2016-06-26 09:51:56.281   0     6
18     x  2016-06-26 09:51:58.279   0     1.998
19     x  2016-06-26 09:52:00.279   0     2

到目前为止，我在 mode 中进行了排序（分组时间和模式），然后找到正确的差异。过境时我无法获得 0。

请指教！

tdm = mydataset.groupby(["Model","Drive Mode"])["timestamp"]
mydataset["DM-Diff"] = tdm.diff().fillna(0) / pd.datetools.timedelta(seconds = 1)
mydataset = mydataset[mydataset["DM-Diff"] < 180]
mydataset = mydataset[mydataset["DM-Diff"] > -1]#hardcoding with range so that it will not include other mode value

Answer 1

你可以使用

In [42]: breaks = df['Mode'].ne(df['Mode'].shift()).cumsum()

In [43]: (df.groupby(breaks)['Timestamp'].diff() / np.timedelta64(1, 's')).fillna(0)
Out[43]:
0     0.000
1     2.003
2     1.997
3     2.000
4     0.000
5     2.000
6     1.999
7     0.000
8     2.000
9     2.001
10    0.000
11    0.000
12    0.000
13    0.000
14    0.000
15    2.000
16    0.000
17    1.998
18    2.000
Name: Timestamp, dtype: float64

---

详情

In [44]: breaks
Out[44]:
0     1
1     1
2     1
3     1
4     2
5     2
6     2
7     3
8     3
9     3
10    4
11    5
12    6
13    7
14    8
15    8
16    9
17    9
18    9
Name: Mode, dtype: int32