https://github.com/shimazadeh/ft_linear_regression

Implementing a modular linear regression from scratch to predict the price of cars using a gradient descent algorithm.
https://github.com/shimazadeh/ft_linear_regression

data-analysis data-science hyperparameter-tuning linear-regression predictive-modeling

Last synced: 7 days ago
JSON representation

Implementing a modular linear regression from scratch to predict the price of cars using a gradient descent algorithm.

• Host: GitHub
• URL: https://github.com/shimazadeh/ft_linear_regression
• Owner: shimazadeh
• Created: 2023-07-07T21:06:54.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-12-26T19:18:37.000Z (11 months ago)
• Last Synced: 2023-12-26T21:02:58.780Z (11 months ago)
• Topics: data-analysis, data-science, hyperparameter-tuning, linear-regression, predictive-modeling
• Language: Python
• Homepage:
• Size: 232 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# DataScience | Linear regression from Scratch | 42Paris

## Objective

To create a modular linear regression model from scratch, train the model on the given dataset, save generated indexes and use them to predict car price depending on it's mileage.

## Requirements

Before running the program, make sure you have the following dependencies installed:

  - numpy: A library for numerical computations.

  - pandas: A library for data manipulation and analysis.

  - matplotlib: A library for data visualization.

  - scikit-learn: A machine learning library for data preprocessing and evaluation.

You can install these packages using pip if you don't have them already: pip install numpy pandas matplotlib scikit-learn

## Usage

- git clone https://github.com/shimazadeh/Ft_linear_regression.git Linear-regression

- cd Linear-regression

- python3 main.py [path/to/dataset.csv] [#iteration] [learning_rate] [mode]

- There are two modes to the program:

  - mode 1: finds the best thetas to be initialized using hyperparameter tuning technique

  - mode 2: uses the parameters from best_params.json file created in the previous option and train and test the model based on that.

## Approach

- Data Preprocessing: The program reads the dataset from the CSV file, normalizes the data to the range of 0-1.

- Train-Test Splitting: The program splits the dataset into training and test sets.

- Hyperparameter Tuning: It uses hyperparameter tuning method to find the best initial parameters (thetas) for the linear regression model.

- Model Training: The linear regression model is trained using gradient descent. The program visualize the training process and print the model parameters, Loss, MSE and MAE each iteration:

![Alt text]()

- Model Evaluation: After training, the program predicts prices using the test set and calculates the Mean Squared Error (MSE) and Mean Absolute Error (MAE) to evaluate the model's performance.

- Visualization: The program visualizes the normalized dataset, the regression model, and the cost function as shown below:

  

![Alt text](output.png)