ft_linear_regression

A simple linear regression project in Python for educational purposes. The goal is to predict the price of a car based on its mileage using linear regression, and to visualize the results and the model's confidence.

Features

• Reads car mileage and price data from CSV datasets.
• Learns optimal parameters (theta0, theta1) using gradient descent.
• Normalizes input data for effective training.
• Exports learned parameters to a CSV file.
• Predicts car prices for any given mileage.
• Calculates R² (coefficient of determination) to indicate model confidence.
• Visualizes the regression line, cost function evolution, and confidence score.

Project Structure

ft_linear_regression/
│
├── datasets/
│   ├── data.csv                # Default dataset
│   ├── big_data.csv
│   ├── negative_data.csv
│   ├── nonlinear_data.csv
│   ├── perfectpositive_data.csv
│   ├── small_data.csv
│   └── variance_data.csv
│
├── model.py                    # Main file: trains, saves, and visualizes the model
├── estimate.py                 # Script to estimate a price given a mileage
├── confidence.py               # Calculates R² confidence score
└── thetas.csv                  # Saved learned parameters after running model.py

Usage

1. Train the Model

Simply run:

python model.py

• Trains the linear regression model on datasets/data.csv.
• Saves the learned parameters (thetas.csv).
• Displays a plot with the data, regression line, confidence score (R²), and cost vs epochs.

2. Estimate Car Price

After training, estimate a car price by mileage:

python estimate.py

• Enter the mileage when prompted.
• Outputs the predicted price.

3. Confidence Score

The R² (confidence) score is printed on the regression plot, indicating how well the model fits the data.

Custom Datasets

You can swap out the dataset by changing the data_path variable in model.py or by providing your own CSV file in the datasets/ folder. The CSV must have columns: km,price.

Requirements

• Python 3.x
• pandas
• numpy
• matplotlib

Install dependencies with:

pip install pandas numpy matplotlib

Notes

• The model uses simple linear regression (1 feature: km).
• Data is normalized for stable and faster learning.