Linear Regression Car Price Predictor

A simple yet effective linear regression model that predicts car prices based on mileage. This project implements gradient descent optimization from scratch to train the model.

Features

• Data Scaling: Normalizes input features to [0, 1] range for better training
• Gradient Descent: Trains the model using batch gradient descent
• Visualization: Plots both scaled and non-scaled regression lines
• CSV Storage: Saves trained parameters for later use
• CLI Flags: Easy control over training and visualization modes

Setup

# Clone/navigate to the project directory
cd ft_linear_regression

# Install dependencies
pip install matplotlib numpy

# Prepare your data
# Ensure you have a Daten.csv file with format:
# mileage,price
# 10000,15000
# 20000,12000
# ...

Usage

Train the model

python train.py

Train and visualize scaled data

python train.py --drawscaled

Train and visualize non-scaled data

python train.py --draw

Make predictions

python predict.py <mileage>

How It Works

1. Data Loading: Reads mileage and price data from Daten.csv
2. Scaling: Normalizes mileage values to [0, 1] range
3. Training: Applies gradient descent to minimize prediction error
4. Visualization: Plots actual data points and the fitted regression line
5. Storage: Saves theta0 and theta1 parameters to theta.csv

Parameters

• Learning Rate: 0.1 (controls step size during training)
• Iterations: 1000 (number of gradient descent steps)

Output

• theta.csv: Stores trained parameters (theta0, theta1, dataset_min, dataset_max)
• Visualization plots: Show regression fit on your data

Project Structure

ft_linear_regression/
├── train.py              # Training script
├── predict.py            # Prediction script
├── Daten.csv            # Input data
├── theta.csv            # Trained parameters
└── README.md            # This file