如何用Pandas与Numpy获取置信度最高的类别及对应置信值？

Hey there! Sounds like you're almost there—you've got the max confidence values locked in, just missing the link to their matching categories. Let's fix that quickly, assuming you're working with a pandas DataFrame (since you used data.max(axis=1)):

Step 1: Target your confidence columns explicitly #

First, define the 3 columns that hold your confidence data. Let's say their names are Cat_X, Cat_Y, Cat_Z—swap these out for your actual column names to avoid accidentally including other columns like TC_Name or Failure.

Step 2: Fetch the category tied to each row's max confidence #

Use pandas' idxmax(axis=1) method—it returns the column name (your category label) where the maximum value occurs in each row. Pair this with your existing max value call:

# Define your confidence columns
confidence_cols = ['Cat_X', 'Cat_Y', 'Cat_Z']

# Add max confidence value to your DataFrame
df['MaxConfidence'] = df[confidence_cols].max(axis=1)

# Add the category name linked to that max value
df['MaxErrCategory'] = df[confidence_cols].idxmax(axis=1)

Step 3: Build your final dataset #

Now just select the 4 columns you need and make a copy to avoid any unintended reference issues:

new_dataset = df[['TC_Name', 'Failure', 'MaxErrCategory', 'MaxConfidence']].copy()

Quick note on tie scenarios #

If multiple categories have the same max confidence in a row, idxmax will return the first one it encounters (left-to-right in your column order). If you need to capture all matching categories for ties, use a custom apply function like this:

def get_all_max_categories(row):
    max_val = row[confidence_cols].max()
    return ', '.join([col for col in confidence_cols if row[col] == max_val])

df['MaxErrCategory'] = df.apply(get_all_max_categories, axis=1)

That should give you exactly the dataset you're aiming for!

内容的提问来源于stack exchange，提问作者Ashutosh Srivastava