根据列和行转换数据框答案

【问题标题】：Transform dataframe based on columns and rows根据列和行转换数据框
【发布时间】：2022-12-17 16:05:02
【问题描述】：

我有以下数据框

CustomerNr	Target	Source	Percentage
1001	A	C	0.2
1004	D	np.nan	0.3
1005	C	D	0.4
1010	A	D	0.5

import numpy as np
df = pd.DataFrame([[1001, 'A','C',0.2], [1004, 'D',np.nan,0.3],[1005, 'C','D',0.4], 
                   [1010, 'A','D',0.5]], columns=['CustomerNr','Target','Source','Percentage'])

到这个（顺便说一下如何制定这个问题的标题）

import numpy as np
df = pd.DataFrame([['1001 Target' , 'A',0.2],
                   ['1001 Source' , 'C',0.2], 
                   ['1004 Target', 'D',0.3],
                   ['1004 Source', np.nan,0.3],
                   ['1005 Target', 'C',0.4],
                   ['1005 Source', 'D',0.4],
                   ['10010 Target', 'A',0.5],
                   ['10010 Source', 'D',0.5],
                  ], columns=['CustomerNr Scope','Value','Percentage'])

CustomerNr Scope	Value	Percentage
1001 Target	A	0.2
1001 Source	C	0.2
1004 Target	D	0.3
1004 Source	NaN	0.3
1005 Target	C	0.4
1005 Source	D	0.4
10010 Target	A	0.5
10010 Source	D	0.5

【问题讨论】：

标签： python dataframe

【解决方案1】：

你可以使用 pandas stack 来实现这个：

(df.set_index(["CustomerNr", "Percentage"])
         .rename_axis("Scope", axis=1)
         .stack(dropna=False)
         .rename("Value")
         .reset_index()
         .assign(CustomerNrScope=lambda df: df[["CustomerNr", "Scope"]].astype(str).apply(" ".join, axis=1)))

或者连接源表和目标表：

df_new = pd.concat([df[["CustomerNr", tscol, "Percentage"]]
               .rename(columns={tscol: "Value"})
               .assign(Scope=tscol)
           for tscol in ["Target", "Source"]])
df_new["CustomerNr Scope"] = df_new.CustomerNr.astype(str) + " " + df_new.Scope


# result
   CustomerNr Value  Percentage   Scope CustomerNr Scope
0        1001     A         0.2  Target      1001 Target
1        1004     D         0.3  Target      1004 Target
2        1005     C         0.4  Target      1005 Target
3        1010     A         0.5  Target      1010 Target
0        1001     C         0.2  Source      1001 Source
1        1004   NaN         0.3  Source      1004 Source
2        1005     D         0.4  Source      1005 Source
3        1010     D         0.5  Source      1010 Source

或者（基于 Zephyrus 的回答）使用带有 Percentage 的 melt 作为额外的 id_var 来立即获得所需的表（假设 Percentage 唯一地依赖于 CustomerNr）：

pd.melt(df, id_vars=['CustomerNr', "Percentage"],
        value_vars=['Target', 'Source'],
        var_name='Scope')

【讨论】：

【解决方案2】：

您可以使用 pandas melt 来取消透视数据框：

df_melted =  pd.melt(df, id_vars=['CustomerNr'], value_vars=['Target', 'Source'], var_name='Scope')

这不包括 'Percentage' 列，但您可以将其合并回新的数据框：

df_melted = df_melted.merge(df[[ 'CustomerNr', 'Percentage']], left_on='CustomerNr', right_on='CustomerNr' )

If you want your `'CustomerNr'` column and `'Scope`' column together you can easily add them together to one column.

【讨论】：

【解决方案3】：

解决方案将是这样的。

import pandas as pd
df = pd.read_csv("test.csv", encoding='utf-8')
df

CustomerNr  Target  Source  Percentage
0   1001    A         C         0.2
1   1004    D         np.nan    0.3
2   1005    C         D         0.4
3   1010    A         D         0.5

解决方案

new_df = pd.DataFrame()
new_indx = len(df)
for ind, row in df.iterrows():
    
    print(new_indx,"  ", ind)
    new_df.at[ind, "CustomerNrScoope"] = str(row['CustomerNr'])+'Target'
    new_df.at[ind,'Value'] = row['Target']
    new_df.at[ind, 'Percentage'] = row['Percentage']
    
    new_df.at[new_indx,"CustomerNrScoope"] = str(row['CustomerNr'])+'Source'
    new_df.at[new_indx,'Value'] = row['Source']
    new_df.at[new_indx,'Percentage'] = row['Percentage']
    new_indx = new_indx +1

输出


CustomerNrScoope    Value   Percentage
0   1001Target        A        0.2
4   1001Source        C        0.2
1   1004Target        D        0.3
5   1004Source        np.nan    0.3
2   1005Target        C        0.4
6   1005Source        D        0.4
3   1010Target        A        0.5
7   1010Source        D        0.5

【讨论】：