Enhancing real estate price prediction using optimized least squares moment balanced machine

Abstract

Real estate price prediction is crucial for urban planning and economic forecasting, as property values are influenced by factors such as location and infrastructure. Traditional methods often struggle to capture the complex relationships between these variables, resulting in inefficient land use and suboptimal resource allocation. To address this challenge, this study introduces the Optimized Least Squares Moment Balanced Machine (OLSMBM), an advanced machine learning model designed to enhance the accuracy of real estate price predictions. The model incorporates key features such as transaction date, house age, proximity to MRT stations, number of convenience stores, and geographic location. The OLSMBM was benchmarked against five other machine learning models, including LSSVM, BPNN, ELSIM, Decision Tree, and Linear Regression. The results from a 10-fold cross-validation demonstrate that OLSMBM consistently outperforms other models across five evaluation metrics, including RMSE (6.977), MAE (4.752), MAPE (13.76%), R (0.846), and R2 (0.728). The comprehensive evaluation, summarized by the Reference Index (RI), showed that OLSMBM achieved a perfect RI score of 1.000, highlighting its superior performance. These findings underscore the potential of the OLSMBM as a decision-support tool, enhancing the accuracy of real estate price predictions and reinforcing data-driven strategies in urban planning.