Question

我正在尝试使用pandas填充缺失值，但无法获得输出。

输入数据：此处缺少某些行值。

             Date_time  current_demand   Temp_Mean  humidity_Mean
0     2018-05-01 00:00         15951.0  300.904267      49.600000
1     2018-05-01 00:15         16075.0  300.904267      49.600000
2     2018-05-01 00:30         15977.0  300.904267      49.600000
3     2018-05-01 00:45         15945.0  300.837600      50.333333
4     2018-05-01 01:00         15868.0  298.889333      59.133333
5     2018-05-01 01:15         15583.0  298.889333      59.133333
6     2018-05-01 01:30         15470.0  298.756000      59.800000
7     2018-05-01 01:45         15301.0  298.756000      59.800000
8     2018-05-01 02:15         14946.0  298.756000      59.800000
9     2018-05-01 02:30         14736.0  298.756000      59.800000
10    2018-05-01 02:45         14630.0  298.502333      59.000000
11    2018-05-01 03:15         14350.0  298.502333      59.000000

我试过的脚本：

import pandas as pd
import numpy as np


df = pd.read_csv(r'submission.csv', index_col=[1], parse_dates=[1], dayfirst=True)

df['Date_time'] = pd.to_datetime(df['Date_time']).dt.time
start = pd.to_datetime(str(df['Date_time'].min()))
end = pd.to_datetime(str(df['Date_time'].max()))
dates = pd.date_range(start=start, end=end, freq='15Min').time


df1 = pd.pivot_table(df, "current_demand", "Temp_Mean", "humidity_Mean").stack(dropna=False).reset_index(name="current_demand")
df1.loc[df1['current_demand'].isnull(), "Temp_Mean", "Temp_Mean" , "humidity_Mean"] = np.nan

浓淡输出：

                Date_time  current_demand   Temp_Mean  humidity_Mean
0     2018-05-01 00:00         15951.0  300.904267      49.600000
1     2018-05-01 00:15         16075.0  300.904267      49.600000
2     2018-05-01 00:30         15977.0  300.904267      49.600000
3     2018-05-01 00:45         15945.0  300.837600      50.333333
4     2018-05-01 01:00         15868.0  298.889333      59.133333
5     2018-05-01 01:15         15583.0  298.889333      59.133333
6     2018-05-01 01:30         15470.0  298.756000      59.800000
7     2018-05-01 01:45         15301.0  298.756000      59.800000
8     2018-05-01 02:00         0        0                 0
9     2018-05-01 02:15         14946.0  298.756000      59.800000
10    2018-05-01 02:30         14736.0  298.756000      59.800000
11    2018-05-01 02:45         14630.0  298.502333      59.000000
12    2018-05-01 03:00         0        0               0
13    2018-05-01 03:15         14350.0  298.502333      59.000000

但是在0的位置 - 由昨天的数据填充（）表示数据或之前数据的前一天）

请建议。提前谢谢

修改

df = df.set_index(['Date_time']).asfreq('15T').ffill()
#df = df.set_index('Date_time').resample('15T').ffill() #as same 
#df = df.asfreq('15T').ffill()

df = df.asfreq('15T').fillna(df.shift(1, freq='d'))

为什么我得到NaN？请让我知道

                     current_demand  Temp_Mean  humidity_Mean
Date_time                                                    
2018-05-01 00:00:00             NaN        NaN            NaN
2018-05-01 00:15:00             NaN        NaN            NaN
2018-05-01 00:30:00             NaN        NaN            NaN
2018-05-01 00:45:00             NaN        NaN            NaN
2018-05-01 01:00:00             NaN        NaN            NaN

Answer 1

使用asfreq或resample进行前向填充：

df = pd.read_csv(r'submission.csv', index_col=[1], parse_dates=[1], dayfirst=True)

df = df.asfreq('15T').ffill()

df = df.resample('15T').ffill()

print (df)

                     current_demand   Temp_Mean  humidity_Mean
Date_time                                                     
2018-05-01 00:00:00         15951.0  300.904267      49.600000
2018-05-01 00:15:00         16075.0  300.904267      49.600000
2018-05-01 00:30:00         15977.0  300.904267      49.600000
2018-05-01 00:45:00         15945.0  300.837600      50.333333
2018-05-01 01:00:00         15868.0  298.889333      59.133333
2018-05-01 01:15:00         15583.0  298.889333      59.133333
2018-05-01 01:30:00         15470.0  298.756000      59.800000
2018-05-01 01:45:00         15301.0  298.756000      59.800000
2018-05-01 02:00:00         15301.0  298.756000      59.800000
2018-05-01 02:15:00         14946.0  298.756000      59.800000
2018-05-01 02:30:00         14736.0  298.756000      59.800000
2018-05-01 02:45:00         14630.0  298.502333      59.000000
2018-05-01 03:00:00         14630.0  298.502333      59.000000
2018-05-01 03:15:00         14350.0  298.502333      59.000000

如果您希望在前几天将NaN替换为fillna并且shift编辑DataFrame，则

df = df.asfreq('15T').fillna(df.shift(1, freq='d'))

Answer 2

使用pd.Grouper和pd.Series.ffill填充前一天数据的空数据的一种方法：

df = pd.DataFrame([['2018-05-01 00:00', 15951.0, 300.904267, 49.600000],
                   ['2018-05-01 00:15', 16075.0, 300.904267, 49.600000],
                   ['2018-05-01 00:30', 15977.0, 300.904267, 49.600000],
                   ['2018-05-01 01:00', 15868.0, 298.889333, 298.889333]],
                  columns=['Date_time', 'current_demand', 'Temp_Mean', 'humidity_Mean'])

df['Date_time'] = pd.to_datetime(df['Date_time'])

grouper = pd.Grouper(key='Date_time', freq='15T')

res = df.groupby(grouper).first().ffill().reset_index()

print(res)

            Date_time  current_demand   Temp_Mean  humidity_Mean
0 2018-05-01 00:00:00         15951.0  300.904267      49.600000
1 2018-05-01 00:15:00         16075.0  300.904267      49.600000
2 2018-05-01 00:30:00         15977.0  300.904267      49.600000
3 2018-05-01 00:45:00         15977.0  300.904267      49.600000
4 2018-05-01 01:00:00         15868.0  298.889333     298.889333

如何使用pandas中的特定填充缺失值（日期和时间）

2 个答案: