About this Website

End-to-end machine learning project for real estate price prediction

Technical Implementation

Data Pipeline

Processed historical HDB transaction data with feature engineering including geocoding, distance calculations to amenities, and temporal features.

Machine Learning

XGBoost gradient boosting model trained on 10+ features with cross-validation for optimal performance.

Backend API

Flask REST API with endpoints for predictions, address lookup, and real-time geocoding via OneMap integration.

Frontend

Interactive single-page application with form validation, auto-fill functionality, and responsive design.

Model Features

The prediction model analyzes the following variables:

  • Location: Town classification with mature/non-mature encoding
  • Property Type: Flat configuration (1-room, 2-room, etc.)
  • Size: Floor area (in sqm)
  • Level: Storey position within the building
  • Lease: Remaining years on the 99-year lease
  • Proximity: Calculated distances to nearest MRT station and shopping mall

Key Features

  • Real-time predictions
  • Intelligent auto-fill from historical database
  • Dynamic geocoding for new addresses
  • Comprehensive input validation
  • RESTful API architecture

⚠️ Disclaimer

Predictions are estimates based on historical patterns and should not be used as the sole basis for financial decisions. Actual property values depend on numerous factors including condition, renovations, market sentiment and individual buyer preferences.