并使用 pandas (v1.3.2) 像这样生成
import pandas as pd
import datetime
input_data = [
["1", datetime.datetime(2023,2,21,20,0,0), 10],
["1", datetime.datetime(2023,2,21,20,30,0), 10],
["2", datetime.datetime(2023,2,21,15,0,0), 15],
["2", datetime.datetime(2023,2,21,15,30,0), 15],
]
df_input = pd.DataFrame(data=input_data, columns=["id", "time", "duration"]).set_index(["id", "time"])
我想根据时隙持续时间(“持续时间”列)“扩展”我的数据帧的第二级(索引列“时间”)。输出数据框应该是这样的:
对第一个 id(“1”)的更多解释:我想要从 20:00 到 20:30 -> 20:00、20:10、20:20、20:30 的所有时隙持续 10 分钟。
我想出了一个解决方案(见下面的代码片段),但它很慢,我想知道是否有更快的内置 pandas 来帮助我处理这个问题。
import pandas as pd
import datetime
input_data = [
["1", datetime.datetime(2023,2,21,20,0,0), 10],
["1", datetime.datetime(2023,2,21,20,30,0), 10],
["2", datetime.datetime(2023,2,21,15,0,0), 15],
["2", datetime.datetime(2023,2,21,15,30,0), 15],
]
df_input = pd.DataFrame(data=input_data, columns=["id", "time", "duration"]).set_index(["id", "time"])
df_output = pd.DataFrame()
for i in range(0, df_input.shape[0], 2):
start_at = df_input.index[i][1]
end_at = df_input.index[i+1][1]
duration = df_input.iloc[i]["duration"]
df_cut = pd.DataFrame(
pd.date_range(
start=start_at,
end=end_at,
freq=f"{duration}min",
).rename("start_at_converted")
)
df_cut["id"] = df_input.index[i][0]
df_cut["duration"] = duration
df_output = pd.concat((df_output, df_cut), axis=0)
df_output = df_output.set_index(["id", "start_at_converted"])
谢谢你的帮助!
groupby.resample
:
freq = {'1': '10min', '2': '15min'}
out = (df_input.reset_index('id').groupby('id')
.apply(lambda g: g.resample(freq[g.name]).ffill())
.drop(columns='id')
# optional, to rename the index
.rename_axis(('id', 'start_at_converted'))
)
输出:
duration
id start_at_converted
1 2023-02-21 20:00:00 10
2023-02-21 20:10:00 10
2023-02-21 20:20:00 10
2023-02-21 20:30:00 10
2 2023-02-21 15:00:00 15
2023-02-21 15:15:00 15
2023-02-21 15:30:00 15
如果你只有一个频率:
out = (df_input.reset_index('id').groupby('id')
.resample('15min').ffill().drop(columns='id')
# optional, to rename the index
#.rename_axis(('id', 'start_at_converted'))
)
输出:
duration
id time
1 2023-02-21 20:00:00 10
2023-02-21 20:15:00 10
2023-02-21 20:30:00 10
2 2023-02-21 15:00:00 15
2023-02-21 15:15:00 15
2023-02-21 15:30:00 15