python - 将数据帧按特定列压缩为包含第一个和最后一个时间戳以及值平均值的行

Question

我已经标记了按时间戳排序的数据组，我想将其减少到开始时间戳和最后时间戳，并获取与该组中该标记相对应的值的平均值。起始数据框示例:

          timestamp          value      mark
1   2016-11-07 11:00:00       0.781726     1  
2   2016-11-07 11:03:00       0.812757     2  
3   2016-11-07 11:05:00       0.845348     2  
4   2016-11-07 11:07:00       0.817394     2  
5   2016-11-07 11:11:00       0.760787     1  
6   2016-11-07 11:13:00       0.807892     1 
7   2016-11-07 11:15:00       0.812965     1  
8   2016-11-07 11:18:00       0.822001     1 

我想要实现的目标:

        start_timestamp         end_timestamp  (mean_)value  mark  
1   2016-11-07 11:00:00   2016-11-07 11:00:00     0.781726      1
2   2016-11-07 11:03:00   2016-11-07 11:07:00     0.825166      2
3   2016-11-07 11:11:00   2016-11-07 11:18:00     0.800911      1

知道执行此操作的最佳方法吗？我应该首先用唯一的标记标记每个批处理吗？

Answer 1

您需要groupby由重复列标记中的唯一组的系列组成，然后aggregate first , last和 mean :

print ((df.mark != df.mark.shift()).cumsum())
1    1
2    2
3    2
4    2
5    3
6    3
7    3
8    3
Name: mark, dtype: int32

df1 = df.groupby((df.mark != df.mark.shift()).cumsum()) \
         .agg({'timestamp': ['first','last'], 'value':'mean', 'mark': 'first'})

#reset MultiIndex in columns
df1.columns = ['_'.join(col) for col in df1.columns]
#if necessary rename columns
df1 = df1.rename(columns=({'timestamp_first':'start_timestamp', 
                           'timestamp_last':'end_timestamp',
                           'mark_first':'mark','value_mean':'(mean_)value'})) \
          .rename_axis(None)

print (df1)         
      start_timestamp       end_timestamp  mark  (mean_)value
1 2016-11-07 11:00:00 2016-11-07 11:00:00     1      0.781726
2 2016-11-07 11:03:00 2016-11-07 11:07:00     2      0.825166
3 2016-11-07 11:11:00 2016-11-07 11:18:00     1      0.800911