在时间序列数据上使用 pd.melt() 旋转数据帧

Pivoting dataframes with pd.melt() on time series data

我这里有一些数据:

       Country/Region  1/22/20  1/23/20  1/24/20  1/25/20  1/26/20  1/27/20
0               Afghanistan        0        0        0        0        0   
1                   Albania        0        0        0        0        0   
2                   Algeria        0        0        0        0        0   
3                   Andorra        0        0        0        0        0   
4                    Angola        0        0        0        0        0   
5       Antigua and Barbuda        0        0        0        0        0   
6                 Argentina        0        0        0        0        0   
7                   Armenia        0        0        0        0        0   
8                 Australia        0        0        0        0        0   
9                 Australia        0        0        0        0        3   
10                Australia        0        0        0        0        0   
11                Australia        0        0        0        0        0   
12                Australia        0        0        0        0        0   
13                Australia        0        0        0        0        0   
14                Australia        0        0        0        0        1   
15                Australia        0        0        0        0        0   
16                  Austria        0        0        0        0        0   
17               Azerbaijan        0        0        0        0        0   
18                  Bahamas        0        0        0        0        0   
19                  Bahrain        0        0        0        0        0   
20               Bangladesh        0        0        0        0        0   

我想重新排列它,使日期成为行,而国家成为列。像这样:

Country/Region   Afghanistan   Albania

1/22/20              0            0
1/23/20              0            0
1/24/20              0            0

等等。我试过使用 pd.melt,但不太清楚如何获得所需的输出。这是我的尝试:

%matplotlib inline
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import math

data = pd.read_csv("covid.csv", sep=",")

data = data.drop(["Province/State","Lat","Long"], axis=1)

data_melted = data.melt(value_vars=data.columns[1:], var_name="Date",value_name="Cases")

          Date  Cases
0      1/22/20      0
1      1/22/20      0
2      1/22/20      0
3      1/22/20      0
4      1/22/20      0
5      1/22/20      0
6      1/22/20      0
7      1/22/20      0
8      1/22/20      0
9      1/22/20      0
10     1/22/20      0
11     1/22/20      0
12     1/22/20      0
13     1/22/20      0
14     1/22/20      0


我也试过:

data_melted = data.melt(value_vars=[data.columns[1:], "Country/Region"])

但这出现了 TypeError: unhashable type: 'Index' 即使 "Country/Region" 不是索引。

在此方面如有任何帮助,我们将不胜感激。

您要转置 table:

df.set_index('Country/Region').T

我注意到Australia重复了很多次,如果你想把它们加起来巩固:

df.set_index('Country/Region').T \
    .groupby(level=0, axis=1) \
    .sum()

如果您指定 pandas.melt. Then, for country columns, run a pivot_table 聚合的 id_vars 参数,那么最初的尝试会奏效,它实际上呈现时间序列数据帧(即 date/time作为索引)用于直接绘图。

data_melted = (data.melt(id_vars = ['Country/Region'],
                         var_name = 'Date', value_name='Cases')
                   .assign(Date = lambda x: pd.to_datetime(x['Date']))

data_pivoted = data_melted.pivot_table(index='Date', columns='Country/Region', 
                                       values='Cases', aggfunc='sum')