Gradient Descent for Multi-Linear Regression

Introduction

This project presents a Python implementation of the gradient descent algorithm for multi-linear regression. Designed to handle problems with 'r' predictors, it allows customization of the learning rate (η) and the number of iteration steps. The implementation is tested on two datasets: advertising.csv and auto.csv, using a suitable train-test split to evaluate the model's performance.

Features

• Implements gradient descent for multi-linear regression problems.
• Customizable learning rate and iteration steps.
• Evaluation using cost function and R-squared test.
• Tested on real-world datasets (advertising.csv and auto.csv).
• Visualization tools for analyzing regression results.

Prerequisites

• Python environment
• Libraries: NumPy, Matplotlib, Seaborn (optional), Scikit-learn.

Installation

Clone the repository to your local machine:

git clone [repository-url]

Usage

1. Navigate to the project directory.
2. Open the provided Jupyter notebooks (main_advertising.ipynb and main_auto.ipynb) to see the implementation on the respective datasets.
3. Modify the parameters (learning rate, iterations, test size) in the Model class instantiation as needed.

Notebooks for Analysis

Two Jupyter notebooks are provided:

1. advertising_analysis.ipynb for the advertising.csv dataset.
2. auto_analysis.ipynb for the auto.csv dataset.

These notebooks guide you through the process of loading the data, creating an instance of the Model class, running the regression analysis, and visualizing the results.