https://github.com/junya737/weighted-pls-regression

A Python implementation of Weighted Partial Least Squares Regression with support for sample weights.
https://github.com/junya737/weighted-pls-regression

machine-learning partial-least-squares-regression scikit-learn

Last synced: 28 days ago
JSON representation

A Python implementation of Weighted Partial Least Squares Regression with support for sample weights.

• Host: GitHub
• URL: https://github.com/junya737/weighted-pls-regression
• Owner: junya737
• License: mit
• Created: 2024-12-27T09:13:43.000Z (29 days ago)
• Default Branch: main
• Last Pushed: 2024-12-27T10:13:21.000Z (29 days ago)
• Last Synced: 2024-12-27T10:29:11.988Z (29 days ago)
• Topics: machine-learning, partial-least-squares-regression, scikit-learn
• Language: Jupyter Notebook
• Homepage:
• Size: 7.81 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# weighted-pls-regression

## Overview

This project implements Weighted Partial Least Squares (WPLS) Regression, a method that incorporates sample weights into the standard PLS regression model.

## Motivation

The default PLSRegression implementation in scikit-learn does not support sample weights. I couldn't also find Python implementations with sample weights.

## Environment

This implementation has been tested in the following environment:

- OS: Ubuntu 22.04.4 LTS

- Python: 3.10.14

- Libraries:

    - numpy: 1.26.4

	- scikit-learn: 1.5.0

## Features

- Supports sample weights for flexible regression modeling.

- Compatible with scikit-learn’s API, enabling integration into pipelines.

## Limitation

- Standardization Toggle: 

Currently, standardization cannot be turned off. 

This feature may be added in future updates.

## Tutorial

A tutorial (tutorial.ipynb) shows how to use this implementation and compares it with scikit-learn’s PLSRegression.

In my environment, the results (MSE, coefficients, intercepts) matched.

## License

This project is licensed under the MIT License.

## Contact

For questions, suggestions, or bug reports, please feel free to:

- Open an issue.

- Email me (junyaihira[@]gmail.com).

Your feedback is highly appreciated!