Forecasting Hong Kong House Prices: An Artificial Neural Network vs Log-linear Regression Approach

Abstract

Modeling the volatility of property prices presents an interesting challenge for researchers. The purpose of the study is to compare an artificial neural network approach and log-linear regression model for predicting private residential property prices in Hong Kong using aggregate variables such real housing prices, real income, interest rate, demographic variables, and so on. The results show that the log-linear regression approach has less the standard error in forecasting. However, an artificial neural network (ANN) has an advantage in its ability to map complicated non-linear relationship between variables and it also has a good predict power.