https://github.com/WillKoehrsen/feature-selector

Feature selector is a tool for dimensionality reduction of machine learning datasets
https://github.com/WillKoehrsen/feature-selector

Last synced: 2 months ago
JSON representation

Feature selector is a tool for dimensionality reduction of machine learning datasets

• Host: GitHub
• URL: https://github.com/WillKoehrsen/feature-selector
• Owner: WillKoehrsen
• License: gpl-3.0
• Created: 2018-06-20T21:14:24.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2024-06-17T22:44:11.000Z (7 months ago)
• Last Synced: 2024-11-11T18:02:54.681Z (2 months ago)
• Language: Jupyter Notebook
• Homepage:
• Size: 4.95 MB
• Stars: 2,229
• Watchers: 97
• Forks: 768
• Open Issues: 42
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-data-science-resources - Feature Selector: Simple Feature Selection in Python
• awesome-data-science-resources - Feature Selector: Simple Feature Selection in Python
• awesome-machine-learning-resources - **[Library - selector?style=social) (Table of Contents)

README

        
# Feature Selector: Simple Feature Selection in Python

Feature selector is a tool for dimensionality reduction of machine learning datasets.

# Methods

There are five methods used to identify features to remove:

1. Missing Values

2. Single Unique Values

3. Collinear Features

4. Zero Importance Features

5. Low Importance Features 

## Usage

Refer to the [Feature Selector Usage notebook](https://github.com/WillKoehrsen/feature-selector/blob/master/Feature%20Selector%20Usage.ipynb) for how to use

## Visualizations

The `FeatureSelector` also includes a number of visualization methods to inspect 

characteristics of a dataset. 

__Correlation Heatmap__

![](images/example_collinear_heatmap.png)

__Most Important Features__

![](images/example_top_feature_importances.png)

Requires:

```

python==3.6+

lightgbm==2.1.1

matplotlib==2.1.2

seaborn==0.8.1

numpy==1.22.0

pandas==0.23.1

scikit-learn==0.19.1

```

## Contact

Any questions can be directed to [email protected]!