Python Pandas：如何添加另一个名称的多索引？

Posted 2023-03-12

【中文标题】Python Pandas：如何添加另一个名称的多索引？

【英文标题】：Python Pandas: How to add another name of multiindex?

【发布时间】：2021-07-28 00:33:06

【问题描述】：

我在玩 Pandas 中的加密数据。合并几个数据框后，我得到了这个

                                            
       timestamp     open      high       low     close   volume    open      high       low      close    volume  
0  1620202740000  54945.31  54987.01  54945.30  54978.49  118.239  54945.31  54987.01  54945.30  54978.49  4345
1  1620202800000  54978.49  55054.00  54972.04  55027.12  337.619  54945.31  54987.01  54945.30  54978.49  134.239   
2  1620202860000  55027.12  55041.05  54950.05  54951.96  131.414  54945.31  54987.01  54945.30  54978.49  14358.239
3  1620202920000  54951.96  55067.36  54951.95  55063.78  176.529  54945.31  54987.01  54945.30  54978.49  1148.239
4  1620202980000  55063.79  55064.00  55000.00  55014.39  107.082  54945.31  54987.01  54945.30  54978.49  18.239

我想在顶部添加另一个级别的索引，所以它会像

                      btc       btc        btc     btc     btc      eth        eth       eth       eth     eth                      
       timestamp     open      high       low     close   volume    open      high       low      close    volume  
0  1620202740000  54945.31  54987.01  54945.30  54978.49  118.239  54945.31  54987.01  54945.30  54978.49  4345
1  1620202800000  54978.49  55054.00  54972.04  55027.12  337.619  54945.31  54987.01  54945.30  54978.49  134.239   
2  1620202860000  55027.12  55041.05  54950.05  54951.96  131.414  54945.31  54987.01  54945.30  54978.49  14358.239
3  1620202920000  54951.96  55067.36  54951.95  55063.78  176.529  54945.31  54987.01  54945.30  54978.49  1148.239
4  1620202980000  55063.79  55064.00  55000.00  55014.39  107.082  54945.31  54987.01  54945.30  54978.49  18.239

所以我很容易像这样添加更多列：

for x in ['btc', 'eth']:
    df.loc[:, (x, 'fast_ema_1min')] = df[x]['close'].rolling(window=1).mean()
    df.loc[:, (x, 'slow_ema_20min')] = df[x]['close'].rolling(window=20).mean()

有人可以建议吗？谢谢。

【参考方案1】：

为了完整起见，如果用expand=True 拆分列，它们将扩展为MultiIndex：

df = df.set_index('timestamp')
df.columns = [pre+col for pre,col in zip(['btc_']*5 + ['eth_']*5, df.columns)]
df.columns = df.columns.str.split('_', expand=True)

#                     btc                                              eth
#                    open      high       low     close   volume      open      high      low     close     volume
# timestamp
# 1620202740000  54945.31  54987.01  54945.30  54978.49  118.239  54945.31  54987.01  54945.3  54978.49   4345.000
# 1620202800000  54978.49  55054.00  54972.04  55027.12  337.619  54945.31  54987.01  54945.3  54978.49    134.239
# 1620202860000  55027.12  55041.05  54950.05  54951.96  131.414  54945.31  54987.01  54945.3  54978.49  14358.239
# 1620202920000  54951.96  55067.36  54951.95  55063.78  176.529  54945.31  54987.01  54945.3  54978.49   1148.239
# 1620202980000  55063.79  55064.00  55000.00  55014.39  107.082  54945.31  54987.01  54945.3  54978.49     18.239

【参考方案2】：

您可以通过以下几种方式创建MultiIndex：

new_columns = pd.MultiIndex.from_arrays([
    (["btc"] * 5) + (["eth"] * 5), 
    df.columns[1:] # exclude "timestamp" from our new columns
])

new_df = df.set_index("timestamp").set_axis(new_columns, axis=1)

print(new_df)
                    btc                                              eth
                   open      high       low     close   volume      open      high      low     close     volume
timestamp
1620202740000  54945.31  54987.01  54945.30  54978.49  118.239  54945.31  54987.01  54945.3  54978.49   4345.000
1620202800000  54978.49  55054.00  54972.04  55027.12  337.619  54945.31  54987.01  54945.3  54978.49    134.239
1620202860000  55027.12  55041.05  54950.05  54951.96  131.414  54945.31  54987.01  54945.3  54978.49  14358.239
1620202920000  54951.96  55067.36  54951.95  55063.78  176.529  54945.31  54987.01  54945.3  54978.49   1148.239
1620202980000  55063.79  55064.00  55000.00  55014.39  107.082  54945.31  54987.01  54945.3  54978.49     18.239

或者，您可以像这样使用MultiIndex.from_product：

new_columns = pd.MultiIndex.from_product([
    ["btc", "eth"], 
    ["open", "high", "low", "close", "volume"]
])

# same as above
new_df = df.set_index("timestamp").set_axis(new_columns, axis=1)

【讨论】：

谢谢@Cameron，它就像一个魅力:)