easyformer 0.1.0

A package for encoding and scaling data for ML pipelines

easyformer

A Python package for creating machine learning pipelines that standardize numerical columns and encode categorical columns using LabelEncoder or OneHotEncoder based on the number of unique values.

Installation

You can install the package from PyPI using pip:

pip install easyformer

Usage

1. Import the Package

from easyformer import create_pipeline
import pandas as pd

2. Load Your Data

# Example: Load a CSV file into a pandas DataFrame
data = pd.read_csv('your_data.csv')

3. Create the Pipeline

pipeline = create_pipeline(data)

4. Fit and Transform Your Data

# Fit the pipeline to your data and transform it
transformed_data = pipeline.fit_transform(data)

5. (Optional) Convert Transformed Data Back to a DataFrame

# If you need the transformed data as a DataFrame
transformed_df = pd.DataFrame(transformed_data, columns=pipeline.named_steps['preprocessor'].get_feature_names_out())
print(transformed_df.head())

Example

Here’s a full example that puts everything together:

from easyformer import create_pipeline
import pandas as pd

# Load your data
data = pd.read_csv('your_data.csv')

# Create the pipeline
pipeline = create_pipeline(data)

# Fit and transform your data
transformed_data = pipeline.fit_transform(data)

# Optionally convert the transformed data back to a DataFrame
transformed_df = pd.DataFrame(transformed_data, columns=pipeline.named_steps['preprocessor'].get_feature_names_out())
print(transformed_df.head())

Description

The easyformer package automates the preprocessing steps for machine learning pipelines by:

• Standardizing Numerical Columns: All numerical columns are scaled using StandardScaler.
• Encoding Categorical Columns:
  • Columns with only two unique values are encoded using LabelEncoder.
  • Columns with more than two unique values are one-hot encoded using OneHotEncoder.
• Removing Original Categorical Columns: The original categorical columns are dropped after encoding.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contributing

If you would like to contribute to this package, please submit a pull request or open an issue on the GitHub repository.