如何将具有特定 id 的行与 pandas 中的前一行合并？

Question

我有一个具有不同值的时间序列数据框：

ID  TimeString            value1  value2   StampDif
0   2021-02-10 17:30:39   0.5     5.2      NaT
1   2021-02-10 17:33:39   0.7     5.5      0 days 00:03:00
2   2021-02-10 17:36:40   0.9     5.5      0 days 00:03:01
3   2021-02-10 17:39:40   0.6     5.4      0 days 00:03:00
4   2021-02-10 17:42:40   0.8     5.0      0 days 00:00:01
.
.
.

现在我想使用均值将 Stampdif 为 1 秒的所有行与前一行合并。我试过：

secdf = df[df["StampDif"] <= pd.Timedelta(1, "sec")]

for idx, row in secdf.iterrows():
    df.iloc[idx-1, dfnanpv.columns != ["TimeString", "StampDif"]] = df.iloc[idx-1:idx+1].mean(axis=0)

但它抛出错误：'Shapes must match', (25,), (2,)

因此，我想在第 3 行中添加以下示例：

ID  TimeString            value1  value2   StampDif
3   2021-02-10 17:39:40   0.7     5.2      0 days 00:03:00

Answer 1

试试这个：

exclude_columns = ['col1', 'col2']
new_df = df.groupby((df['StampDif'] > pd.Timedelta(1, 'second')).cumsum()).agg({col: 'mean' for col in df.columns.difference(exclude_columns)}).reset_index(drop=True)

输出：

>>> new_df
           TimeString  value1  value2
0 2021-02-10 17:30:39     0.5     5.2
1 2021-02-10 17:33:39     0.7     5.5
2 2021-02-10 17:36:40     0.9     5.5
3 2021-02-10 17:41:10     0.7     5.2

如何将具有特定 id 的行与 pandas 中的前一行合并？

How to merge rows with a specific id with the previous row in pandas?

python

pandas

time-series