Pandas将分类值转换为数值并生成新列（非LabelEncode）

Hey there! Let's walk through this step by step—since you're new to Python and pandas, I'll keep things clear and avoid overly complex code. Here's how you can generate new _scl columns without modifying your original data:

Step 1: Identify Columns with Ordered Categorical Values #

First, we need to pick out which columns have the ordered string/categorical values you want to convert. There are two common scenarios here:

• Your columns are already marked as ordered categorical in pandas
• They're stored as object (string) type but represent an ordered set (like "low" → "medium" → "high")

Code to Filter Columns #

import pandas as pd

# Assume your original DataFrame is named `df`
# Option 1: Auto-detect existing ordered categorical columns
ordered_cat_cols = [
    col for col in df.select_dtypes(include='category').columns
    if df[col].cat.ordered
]

# Option 2: Manually specify columns if you know which ones are ordered
# (Use this if your columns are still string/object type)
# ordered_cat_cols = ["your_column_name_1", "your_column_name_2"]

Step 2: Generate New _scl Columns with Numeric Values #

Now we'll loop through each identified column and create a new column (with _scl suffix) that holds the numeric equivalent of the ordered values. We'll make sure not to modify the original columns.

Code for Conversion #

# If you want to keep your original DataFrame untouched, make a copy first
new_df = df.copy()

# Loop through each ordered column
for col in ordered_cat_cols:
    if new_df[col].dtype == 'category':
        # For existing ordered categories, use .cat.codes to get numeric values
        new_df[f"{col}_scl"] = new_df[col].cat.codes
    else:
        # For string/object columns, define your ordered mapping first
        # Replace this with your actual value-to-number mapping
        order_mapping = {
            "low": 0,
            "medium": 1,
            "high": 2
        }
        # Map the strings to numbers and create the new column
        new_df[f"{col}_scl"] = new_df[col].map(order_mapping)
        
        # Alternative (more robust): Convert to ordered category first, then get codes
        # new_df[f"{col}_scl"] = pd.Categorical(
        #     new_df[col],
        #     categories=["low", "medium", "high"],  # Match your order
        #     ordered=True
        # ).codes

Step 3: Verify Your New Columns #

Check that the new _scl columns were created correctly:

# View the first few rows of all _scl columns
print(new_df.filter(like='_scl').head())

Key Notes for You #

• Adjust the mapping: The order_mapping dictionary needs to match your actual string values and their desired numeric order. For example, if a column uses "beginner" → "intermediate" → "expert", update the mapping accordingly.
• 81 columns isn't a problem: The loop will handle as many ordered columns as you identify—no need to write code for each one individually.
• Original data stays intact: By working on new_df (a copy of your original DataFrame), your original columns won't be modified at all.

内容的提问来源于stack exchange，提问作者CGermain