如何简洁扁平化Pandas Groupby聚合后的多级索引列？

Great question! Manually renaming columns after groupby-aggregation is tedious and error-prone, especially as your aggregation logic grows. Let's cover two straightforward approaches to avoid or flatten those multi-level columns automatically:

1. Define Column Names Directly in agg() (Best Practice) #

Instead of letting Pandas generate multi-level columns, you can specify custom column names while aggregating using tuple syntax. This skips the multi-level index entirely:

import pandas as pd
import numpy as np

df = pd.DataFrame( {'A': [1,1,1,2,2,2,3,3,3], 'B': np.random.random(9), 'C': np.random.random(9)} )

# Specify column names alongside aggregation functions
out = df.groupby('A').agg(
    B_mean=('B', np.mean),
    B_std=('B', np.std),
    C_median=('C', np.median)
)

print(out)

This gives you the flattened columns directly, no post-processing needed:

B_mean    B_std  C_median
A
1  0.791846  0.091657  0.394167
2  0.156290  0.202142  0.453871
3  0.482282  0.382391  0.892514

2. Flatten Existing Multi-Level Columns #

If you already have a DataFrame with multi-level columns (like your original example), you can automate renaming by joining the level values:

Option A: List Comprehension with str.join #

# Using your original out DataFrame with multi-level columns
out.columns = ['_'.join(col) for col in out.columns]

Option B: Use Pandas' get_level_values #

This is useful if you want more control over how levels are combined:

out.columns = out.columns.get_level_values(0) + '_' + out.columns.get_level_values(1)

Both methods will produce the same flattened column names as your manual approach, but without the hassle of typing each name manually.

Bonus: Save to Text File #

Once your columns are flattened, saving to a text file (like CSV or tab-separated) is straightforward:

out.to_csv('aggregated_data.txt', sep='\t') # Use sep=',' for standard CSV format

内容的提问来源于stack exchange，提问作者Haleemur Ali