Forecasting House Price Index using artificial neural network technique

Abstract

Property forecasting is an important component within the decision-making process for investors & governments in supporting asset allocation, formulating property funding strategies, and determining suitable policies. Moreover, information on potential and risks provides Multiple Regression Analysis has been accepted as the most common method in property forecasting technique. Yet the problem associated with this technique arises from the limited ability of the model to effectively deal with relations among variables, nonlinearity, and multicolinearity. Therefore, several alternative techniques had been explained to improve the reliabilities of forecasting information, one of the most commonly accepted techniques is the application of artificial neural network. Consequently this work intent to develop a forecasting model for residential property prices in Malaysia using an artificial neural network approach. The objective is tested with real time market data. Unemployment rate, population, mortgage rate and household income are chosen as the housing influencing variables (independent variable). On the other hand, Housing Price Index (HPI) was chosen as the dependent variable due to its ability to reflect the country price trend. Quarterly time-series data from 2000 until 2009 are obtained and applied for training and testing the ANN model. Subsequently, data for 2010 and 2011 is used to validate the model. Model validation gives Mean Absolute Percentage Error (MAPE) of 8% thereby suggesting high accuracy of adopting neural network in ability to learn, generalize, and converge time series data efficiently and produce reliable forecasting information.