README.md

Feature Selection Using Wrapper Methods

Project Brief

Objective: In this project, I will perform feature selection using three different wrapper methods on a dataset obtained from a survey about eating habits and weight. The goal is to identify a smaller subset of features that can accurately predict whether a survey respondent is obese or not.

Data source: UCI Machine Learning Repository - Estimation of obesity levels

Tasks Overview

Here's a breakdown of the tasks I'll be performing using Python and various packages:

1. Data Loading and Inspection: I'll load the dataset and inspect its structure using pandas. This will help me understand the available features and the target variable.

2. Logistic Regression Model: I'll create a logistic regression model using LogisticRegression from sklearn to establish a baseline accuracy for comparison.

3. Sequential Forward Selection (SFS): I'll implement SFS using the SequentialFeatureSelector class from mlxtend to select a subset of features that maximizes accuracy. I'll then evaluate the model's accuracy on this reduced feature set.

4. Sequential Backward Selection (SBS): I'll implement SBS using the SequentialFeatureSelector class from mlxtend to select another subset of features that maximizes accuracy. Again, I'll evaluate the model's accuracy on this subset.

5. Recursive Feature Elimination (RFE): I'll standardize the data and apply RFE using RFE from sklearn to select a subset of features. I'll also evaluate the model's accuracy on this reduced feature set.

6. Visualizing Results: I'll visualize the results of SFS and SBS by plotting the model accuracy as a function of the number of features used.