如何从列名字符串创建数据帧索引？奖励问题：然后获取传播数据框

Question

我有一个包含 1300 行和 400 列的数据框，如下所示

            AOC.2017Jan  AOC.2017Feb  ...  ZTP.2021Oct  ZTP.2021Nov
VALUE_TIME                            ...                          
2016-07-07          NaN          NaN  ...          NaN          NaN
                ...          ...  ...          ...          ...
2021-10-14          NaN          NaN  ...          NaN     101.1000
2021-10-15          NaN          NaN  ...          NaN      88.6250
2021-10-18          NaN          NaN  ...          NaN      90.1375
2021-10-19          NaN          NaN  ...          NaN      91.1125
2021-10-20          NaN          NaN  ...          NaN      93.5500

我想检索3个字母和日期作为索引（长格式），最Pythonic/Pandas的方法是什么？

这大致是预期的效果，前 2 个日期之间的行未显示

VALUE_TIME Group Date      Value                        
2016-07-07 AOC   2017Jan   NaN   
           AOC   2017Feb   NaN 
           ZTP   2021Oct   NaN
           ZTP   2021Nov   NaN
2021-10-14 ZTP   2021Nov  101.1000
2021-10-15 ZTP   2021Nov  88.6250
2021-10-18 ZTP   2021Nov  90.1375
2021-10-19 ZTP   2021Nov  91.1125
2021-10-20 ZTP   2021Nov  93.5500

顺便说一句，我的最终目标是拥有一个数据框，显示具有相同 value_time 和日期

的任何产品（AOC、ZTP 等）对之间的所有可能价差

最终目标应该是这样的

VALUE_TIME Spread    Date      Value                        
2016-07-07 AOC-BBC   2017Jan   xxx
           AOC-BBC   2017Feb   xxx
           AOC-ZTP   2017Jan   xxx
           AOC-ZTP   2017Feb   NaN 
           BBC-ZTP   2017Feb   NaN 
           BBC-ZTP   2017Feb   NaN 
2016-07-08 AOC-BBC   2017Jan   xxx
           AOC-BBC   2017Feb   xxx
           AOC-ZTP   2017Jan   xxx
           AOC-ZTP   2017Feb   NaN 
           BBC-ZTP   2017Feb   NaN 
           BBC-ZTP   2017Feb   NaN

Answer 1

回答你的第一个问题：

df.columns=pd.MultiIndex.from_tuples([x.split('.') for x in df], names=['Group','Date'])

df.stack(level=[0,1], dropna=False).to_frame(name='Value')

你会得到这样的东西：

                          Value
VALUE_TIME Group Date          
2016-07-07 AOC   2017Feb    NaN
                 2017Jan    NaN
                 2021Nov    NaN
                 2021Oct    NaN
           ZTP   2017Feb    NaN
                 2017Jan    NaN
                 2021Nov    NaN
                 2021Oct    NaN
2021-10-14 AOC   2017Feb    NaN
                 2017Jan    NaN
                 2021Nov    NaN
                 2021Oct    NaN
           ZTP   2017Feb    NaN
                 2017Jan    NaN
                 2021Nov  101.1
                 2021Oct    NaN

但是你的另一个问题不是很清楚。

如何从列名字符串创建数据帧索引？奖励问题：然后获取传播数据框

how to make dataframe index from column name string? bonus question: then to get the spread dataframe

python

pandas

dataframe

multi-index