Predicting Price of Taiwan Real Estates By Neural Networks and Support Vector Regression

摘要

The main purpose of this study is to predict the real estate price in Taiwan efficiently. Neural networks and Support Vector Regression are applied and compared. Variables are first selected from previous research and than chose by stepwise procedure and trial-and-error methods. It is found that SVR with trial-and-error method performed the best with MAPE=4.466% and R~2=0.8540. In addition, Rediscount rate, Money supply, and Price of last month are the three common variables for both BPNN and SVR.