Boosted Tree Ensembles for Artificial Intelligence Based Automated Valuation Models (AI-AVM)

Area of Science:

• Artificial Intelligence
• Machine Learning
• Real Estate Economics

Background:

• Accurate housing price prediction is crucial for market stability and investment decisions.
• Existing valuation models may lack the precision needed for dynamic real estate markets.
• The Singapore housing market presents unique characteristics influencing property values.

Purpose of the Study:

• To develop and evaluate an artificial intelligence based automated valuation model (AI-AVM) for predicting Singapore housing prices.
• To compare the predictive performance of a boosting tree ensemble model against traditional methods.
• To assess the accuracy and robustness of the AI-AVM for both public and private housing segments.

Main Methods:

• Utilized a boosting tree ensemble technique for the AI-AVM.
• Trained models on over 300,000 Singapore housing transactions (1995-2017).
• Compared AI-AVM performance with decision tree and multiple regression analysis (MRA) models.

Main Results:

• The boosting AI-AVM demonstrated superior accuracy and robustness compared to other models.
• The model explained 91.33% (public) and 94.28% (private) of price variances.
• Achieved low mean absolute percentage errors: 8.55% (public) and 5.34% (private).
• Out-of-sample forecasting for 2018 sales showed prediction errors between 5% and 9%.

Conclusions:

• The developed AI-AVM is a highly effective tool for predicting Singapore housing prices.
• Boosting tree ensemble technique provides a robust and accurate approach for real estate valuation.
• The AI-AVM shows strong predictive power for both public and private housing markets.