A SVR based forecasting approach for real estate price prediction

The support vector machine (SVM), proposed by Vapnik (1995), has been successfully applied to classification, cluster, and forecast. This study proposes support vector regression (SVR) to forecast real estate prices in China. The aim of this paper is to examine the feasibility of SVR in real estate price prediction. To achieve the aim, five indicators are selected as the input variables and real estate price is used as output variable of the SVR. The quarterly data during 1998–2008 are employed as the data set to construct the SVR model. With the scenarios, real estate prices in future are forecasted and analyzed. The forecasting performance of SVR model was also compared with BPNN model. The experimental results demonstrate that based on the mean absolute error (MAE), the mean absolute percentage error (MAPE) and the root mean squared error (RMSE), the SVR model outperforms the BPNN model and the SVR based approach was an efficient tool to forecast real estate prices.