Pandas透视表格式不符，已尝试reset_index仍未解决求助

It sounds like your pivot table is returning a hierarchical column structure (a MultiIndex) that’s not matching your expected output—even after calling reset_index(). Let’s walk through a few practical fixes to get your DataFrame looking how you want it.

1. Flatten Hierarchical Column Names #

When you use the columns parameter in pivot_table, pandas creates a two-level column index: one for your value (VWR_CLICKTHROUGH) and another for each unique CLICK_TAG value. Resetting the row index doesn’t touch this column structure, so you’ll need to flatten it manually:

# After creating your pivot table and resetting the index
intractionTableClicks.columns = ['_'.join(col).strip() for col in intractionTableClicks.columns.values]

This combines the nested column levels into a single, readable string (e.g., VWR_CLICKTHROUGH_ClickTag_Name instead of a stacked header).

2. Restructure the Pivot Table to Avoid Hierarchical Columns #

If you actually want CLICK_TAG to be a row-level attribute instead of a column, adjust your pivot table definition by moving CLICK_TAG to the index instead of the columns:

# Alternative pivot table setup for a flat structure
intractionTableClicks = pd.pivot_table(
    read_sql_preroll_interaction,
    index=["PLACEMENT", "CLICK_TAG"],
    values=["VWR_CLICKTHROUGH"],
    aggfunc=np.sum,
    fill_value=0
).reset_index()

This gives you a simple DataFrame where each row represents a unique combination of PLACEMENT and CLICK_TAG, with a single column for the summed click-throughs.

3. Diagnose the Current Structure #

If you’re still unsure what’s off, print the current structure to pinpoint the issue:

print(intractionTableClicks.head())
print(intractionTableClicks.columns)

This will show you the exact column hierarchy and sample data rows, helping you confirm whether the problem is with nested columns, incorrect grouping, or something else.

Try these steps—flattening the columns or adjusting the index/columns parameters should resolve your formatting issue.

内容的提问来源于stack exchange，提问作者abhilash mishra