如何根據另一列的最大值對一列進行重采樣？

Question

比如說，我有一個具有不規則時間步長的地表風數據集，如下所示。 實際數據集有許多其他列和數千行。

Time, Speed, Direction
2023-1-1 01:00:00, 6, 90
2023-1-1 02:00:00, 6, 70
2023-1-1 03:00:00, 9, 70
2023-1-1 04:00:00, 6, 230
2023-1-1 06:00:00, 2, 320
2023-1-1 08:00:00, 2, 100
2023-1-1 11:00:00, 3, 140
2023-1-1 15:00:00, 12, 10
2023-1-1 16:00:00, 13, 20
2023-1-1 17:00:00, 15, 60
2023-1-1 18:00:00, 10, 80

我想通過計算最大速度和與最大速度對應的方向，將數據重新采樣為固定時間步長 (00 03 06 09 12 15 18 21)。 如何才能做到這一點？ 我正在弄清楚這樣的事情，但它不起作用。

df3h = df.resample('3H').agg({ # 3H Does not work if the time series donot start at 00:00
'Speed':'max'
'Direction': lambda x, x.loc[x.Speed.idxmax(),'Direction'] # This Won't Work!
})

Answer 1

您可以在創建新的 dataframe 之前使用groupby和idxmax來完成。首先在時間列上使用floor來獲取 3h 的組。

# create 3h bins
_3h = df['Time'].dt.floor('3H')
print(_3h) # if Time is index, then do df.index.floor('3H')
# 0    2023-01-01 00:00:00
# 1    2023-01-01 00:00:00
# 2    2023-01-01 03:00:00
# 3    2023-01-01 03:00:00
# 4    2023-01-01 06:00:00
# ...

#use it to groupby and get the index of the max Speed with idxmax
_idxmax = df.groupby(_3h)['Speed'].idxmax()
print(_idxmax)
# Time
# 2023-01-01 00:00:00     0
# 2023-01-01 03:00:00     2
# 2023-01-01 06:00:00     4
# ...

# create the result dataframe
new_df = (
    df.loc[_idxmax, ['Speed','Direction']]
      .set_index(_idxmax.index)
      # in case a 3h bin is missing like 12:00 in your example
      .reindex(pd.Index(
          pd.date_range(_3h.min(), _3h.max(), freq='3h'), 
          name='Time'))
      # if you want Time as column
      .reset_index()
)
print(new_df)
#                  Time  Speed  Direction
# 0 2023-01-01 00:00:00    6.0       90.0
# 1 2023-01-01 03:00:00    9.0       70.0
# 2 2023-01-01 06:00:00    2.0      320.0
# 3 2023-01-01 09:00:00    3.0      140.0
# 4 2023-01-01 12:00:00    NaN        NaN
# 5 2023-01-01 15:00:00   15.0       60.0
# 6 2023-01-01 18:00:00   10.0       80.0