A focused regression notebook comparing categorical encoding choices for used-car price estimation.

The notebook loads Ford listings, explores numeric/categorical features, scales year, mileage, tax, MPG, and engine size, then fits linear regression models over encoded feature matrices. Evaluation uses a held-out split and R².

Experiment path

ford.csv → inspect/clean → encode categorical variables → standardize numerics
         → train/test split → linear regression → R² comparison

Run

git clone https://github.com/ReaperXD67/Linear-Regression-.git
cd Linear-Regression-
python -m venv .venv
pip install -r requirements.txt
jupyter notebook Untitled.ipynb

The dataset is already present, so the Kaggle download cell is optional. If used, configure Kaggle credentials outside the repository.

Interpretation notes

• R² measures held-out variance explained; it does not guarantee accurate prices for every segment.
• Random splitting can overstate real deployment performance when market conditions change over time.
• A production evaluation should add MAE/RMSE, residual slices, temporal validation, feature provenance, and prediction intervals.