🏠 Ames Housing Price Prediction

🎯 Objective

This project aims to predict house prices in Ames, Iowa, based on a dataset with 79 explanatory variables.
It demonstrates end-to-end data science workflow: from data exploration to feature engineering, model training, evaluation, and stacking.
On Kaggle, this solution ranked in the Top 25% Leaderboard.

---

📊 Dataset

• Source: Kaggle Ames Housing Dataset
• 1460 observations, 79 features (categorical, ordinal, and continuous).
• Target: SalePrice.

---

⚙️ Pipeline

1. Exploratory Data Analysis (EDA) — distributions, correlations, missing values.
2. Feature Engineering — handling NAs, log transforms, encoding categorical variables, scaling.
3. Modeling — regression models (Linear Regression, Ridge, Lasso, Random Forest, XGBoost, LightGBM).
4. Stacking — ensemble with Ridge as meta-model.
5. Evaluation — RMSE, R², cross-validation.

---

🚀 Results

• Baseline (Linear Regression): RMSE ~0.21 (log error).
• Tree-based models (XGBoost, LightGBM): improved performance significantly.
• Final stacking (Ridge meta-model): Top 25% Kaggle Leaderboard.

---

📂 Repository structure

kaggle-house/ ├── README.md ├── requirements.txt ├── LICENSE ├── .gitignore │ ├── data/ │ ├── raw/ # Original Kaggle dataset │ ├── processed/ # Cleaned & feature-engineered dataset │ ├── notebooks/ │ ├── 01_EDA.ipynb │ ├── 02_FeatureEng.ipynb │ ├── 03_Modeling.ipynb │ ├── 04_Evaluation.ipynb │ └── 05_Stacking.ipynb │ ├── src/ │ ├── data_prep.py │ ├── train.py │ └── evaluate.py │ ├── models/ │ ├── baseline_model.pkl │ ├── final_model.pkl │ └── model_card.md │ └── reports/ └── figures/

---

🛠️ How to run

Install dependencies

git clone https://github.com/driksey/ames-housing-price-prediction.git
cd ames-housing-price-prediction
pip install -r requirements.txt