Skip to content

Latest commit

 

History

History
62 lines (39 loc) · 3.42 KB

README.md

File metadata and controls

62 lines (39 loc) · 3.42 KB
Raw

Feature Selection for Machine Learning

This repository contains the code for three main methods in Machine Learning for Feature Selection i.e. Filter Methods, Wrapper Methods and Embedded Methods. All code is written in Python 3.

Status: Ongoing

Requirements

1. Python 3.5 +

2. Jupyter Notebook

3. Scikit-Learn

4. Numpy [+mkl for Windows]

5. Pandas

6. Matplotlib

7. Seaborn

8. mlxtend

Datasets

1. Santander Customer Satisfaction Dataset

2. BNP Paribas Cardif Claims Management Dataset

3. Titanic Disaster Dataset

4. Housing Prices Dataset

Filter Methods

S.No. Name About Status
1. Constant Feature Elimination This notebook explains how to remove the constant features during pre-processing step. Completed
2. Quasi-Constant Feature Elimination This notebook explains how to get the Quasi-Constant features and remove them during pre-processing. Completed
3. Duplicate Features Elimination This notebook explains how to find the duplicate features in a dataset and remove them. Completed
4. Correlation This notebook explains how to get the correlation between features and between features and target and choose the best features. Completed
5. Machine Learning Pipeline This notebook explains how to use all the above methods in a ML pipeline with performance comparison. Completed
6. Mutual Information This notebook explains the concept of Mutual Information using classification and Regression to find the best features from a dataset. Completed
7. Fisher Score Chi Square This notebook explains the concept of Fisher Score chi2 for feature selection. Completed
8. Univariate Feature Selection This notebook explains the concept of Univariate Feature Selection using Classification and Regression. Completed
9. Univariate ROC/AUC/MSE This notebook explains the concept of Univariate Feature Selection using ROC AUC scoring. Completed
10. Combining all Methods This notebook compares the combined performance of all methods explained. Completed

Wrapper Methods

S.No. Name About Status
1. Step Forward Feature Selection This notebook explains the concept of Step Forward Feature Selection. Completed
2. Step Backward Feature Selection This notebook explains the concept of Step Backward Feature Selection. Completed
3. Exhaustive Search Feature Selection This notebook explains the concept of Exhaustive Search Feature Selection. Completed

Embedded Methods

S.No. Name About Status