🏡 House Price Prediction Challenge This repository contains a complete end-to-end machine learning workflow to solve a real-world House Price Prediction problem using a dataset from Kaggle. The project focuses on data wrangling, exploratory data analysis, feature engineering, model building, hyperparameter tuning, and model evaluation.
📊 Project Overview
The goal is to predict house prices (in Lakhs) based on several property-related features, such as square footage, number of rooms, location, and more. A Random Forest Regressor was trained using an optimized pipeline, and the best model was selected through GridSearchCV.
🔍 Key Highlights 📥 Dataset downloaded using kagglehub
🧹 Custom wrangle() function for preprocessing and cleaning
🧠 Baseline model with DummyRegressor
⚙️ Full pipeline: imputation → scaling → encoding → regression
🧪 Cross-validation and hyperparameter tuning with GridSearchCV
📈 Evaluation using RMSE and R² metrics
🗺️ Feature importance analysis and visualization
📦 Model serialization using joblib
🧾 Reusable prediction function for new data
🛠️ Technologies Used Python (3)
Pandas, NumPy, Seaborn, Matplotlib, Plotly
Scikit-learn (Pipeline, RandomForestRegressor, GridSearchCV)
Scipy (for statistical tests)
Joblib (for model persistence)
📈 Results Best cross-validated RMSE: ~19.36
Test RMSE: ~19.54
Test R²: ~0.76
Most important features: SQUARE_FT, LONGITUDE, LATITUDE, and location-related encodings
📌 Future Work Try advanced models (e.g., XGBoost, LightGBM)
Deploy the model as an API (e.g., with FastAPI or Flask)
Integrate with streamlit for interactive dashboard
📬 Contact Feel free to open an issue or reach out if you have suggestions or questions!