Predicting Residential Housing Rent Prices Using Machine Learning and Statistical Modelling Approaches

Abstract

Accurate estimation of residential rental prices is critical for property valuation, investment decision-making, and housing market analysis in urban areas. The rapid urban development in Selangor and Kuala Lumpur has heightened the complexity of determining rental prices, thereby necessitating data-driven analytical approaches. This study identifies key factors influencing residential rental prices and evaluates the predictive performance of statistical and machine learning models using housing data from Selangor and Kuala Lumpur. Stepwise regression was initially applied to identify significant explanatory variables affecting monthly rental prices. Multiple Linear Regression served as the baseline statistical model, while Random Forest and Extreme Gradient Boosting (XGBoost) were implemented to enhance predictive accuracy through nonlinear learning. The dataset was partitioned into training (80%) and testing (20%) subsets, and model performance was assessed using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). The results show that property size, furnishing status, location, and completion year are among the most influential factors affecting rental prices. In terms of predictive performance, machine learning models outperformed the traditional regression approach, with XGBoost achieving the lowest test-set prediction error, followed by the Random Forest. These findings demonstrate the effectiveness of machine learning techniques for predicting residential rental prices and offer valuable insights for property analysts, policymakers, and real estate practitioners in Malaysia.