UrbanNest — Intelligent Flat Price Estimator for Kolkata
Project Description
UrbanNest is a Streamlit-powered machine learning web application that estimates residential flat prices in the Kolkata region. Built using advanced regression models trained on 1,100+ real estate listings scraped from 99acres, incorporating structural, amenity-based, and spatial features.
Features
- Web-based UI with real-time price prediction and model selection
- Reverse geocoding support for location feature extraction
- Feature contribution plots for explainability
- Input validation and robust feature handling
- Joblib-based model serialization for efficient loading
Model Performance
- XGBoost: RMSE = ₹30.30L, R² = 0.8425 — best model
- Gradient Boosting: RMSE = ₹31.07L, R² = 0.8344
- Random Forest: RMSE = ₹35.48L, R² = 0.7840
- ElasticNet: RMSE = ₹36.35L, R² = 0.7733
Technical Stack
- Language: Python
- Frameworks: Streamlit, scikit-learn, XGBoost
- Libraries: Pandas, NumPy, joblib, geopy
- Tools: Google Colab, VS Code
How It Works
- Web scraping of flat listings across Kolkata (1,100+ entries).
- Feature engineering: BHK, area, floors, amenities, geo-coordinates.
- Training and evaluation of multiple regression models.
- Streamlit interface with real-time prediction and feature contribution visualization.