Linear Regression

This repository contains Linear Regression implementation from scratch.

About

This implementation contains training with batch and mini-batch gradient descent. To reproduce the results random_state parameter can be used.

Implemented methods of regularization:

• L1 (Lasso),
• L2 (Ridge),
• ElasticNet,

Custom learning rate (float or callable per epoch) can be set during model initialization.

Implemented and supported metrics:

• MSE,
• RMSE,
• MAE,
• MAPE,
• R² score,

Verbose training output can be set for every n epochs.

Dependencies

To install all required dependencies, execute the following command:

pip install requirements.txt

Usage

To start main script, execute the following command:

python main.py

Run Linear Regression for House Pricing Prediction

The notebook notebooks/kaggle_example.ipynb demonstrates how to use the implemented from scratch Linear Regression on a real-world dataset.

1. First, you need to download dataset from Kaggle: House Pricing Prediction.
2. Create data/ directory in the root of the project, rename the downloaded file to "housing.csv" and put there.
3. Go to notebooks/ directory and run housing.ipynb notebook.