ADVANCING MALAYSIAN HOUSE PRICE INDEX PREDICTIONS: THE POTENTIAL OF ARTIFICIAL NEURAL NETWORK

This study develops a localized house price index by employing an Artificial Neural Network (ANN) forecasting model to enhance the organization and efficiency of the Malaysian property market. Traditional Malaysian House Price Index (MHPI), based on a hedonic model, exhibits multicollinearity among house attributes, limiting its ability to accurately capture house price trends. While MHPI provides aggregate indices, the construction of disaggregate indices remains underexplored, potentially reducing the effectiveness of local housing market analysis and long-term planning. This research aims to identify key house attributes influencing prices and to evaluate the capability of ANN in forecasting the Johor Bahru house price index. The study applies ANN to disaggregate transaction data of double-story terrace houses in Taman Mutiara Rini and Taman Bukit Indah, comparing actual and forecasted prices using R-squared (R²), Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and Mean Absolute Deviation (MAD). Findings reveal significant differences between forecasted and actual indices, demonstrating ANN’s superior ability to capture local price variations. The study highlights ANN as a robust tool for constructing disaggregate house price indices, offering improved insight and efficiency for monitoring and planning within local housing markets in Malaysia.