Real Estate Price Analytics & Prediction

Overview

Forecast Singapore HDB resale flat prices using transaction history and geospatial features.

Data Sources

• Data.gov.sg: 174,893 transactions (160,858 train / 14,035 test; 28 columns)
• OneMap API: Coordinates for flats, MRT stations, elite schools
• Kaggle: Supplementary amenity and demographic data

Data Processing & EDA

• Notebooks:
  1. 1_feature_engineering+EDA.ipynb – ingestion, cleaning, Geopy distance calculations, and exploratory analysis (mean floor area 107.23 sqm; mean price SGD 493 000).
  2. 2_model_building.ipynb – feature encoding (197 total), model training and evaluation.

Modeling

• Train/test split: 160,858 / 14,035 records
• Models & Test Performance:
  • Linear Regression: R² = 0.864, MAE = SGD 48 455
  • Decision Tree: R² = 0.808, MAE = SGD 54 335
  • Random Forest: R² = 0.913, MAE = SGD 35 874
  • XGBoost: R² = 0.9537, MAE = SGD 27 568, RMSE = SGD 38 274, MAPE = 4.66%