Artificial neural network model (ANN) for Malaysian housing market analysis

Abstract

The Hedonic Model, a traditional method for forecasting house prices has been criticised due to nonlinearity, multicollinearity and heteroskedasticity problems, which were argued to affect estimation accuracy. Unlike the Hedonic Model, the Artificial Neural Network Model (ANN), permits nonlinear relationships and thus avoids the problems plaguing the Hedonic Model resulting in superior forecasting performance. Despite these advantages, attempts to model house prices using ANN are limited in geography and data thus besetting the usefulness of the results. To address the research gap, this paper aims to establish such a new model using ANN in forecasting house prices. A sample of double-storey terraced houses transacted in Johor Bahru are analysed using ANN. The findings illustrate a superior forecasting performance for ANN through high values of goodness of fit and low values of errors. This paper adds to the house price modelling literature and provides new knowledge to both academics and practitioners.