Housing Market Hedonic Price Study Based on Boosting Regression Tree

Abstract

Based on the purchase price data of new real estate markets three cities in China, Beijing, Shanghai, and Guangzhou, including architectural features, neighborhood property features, and location features, in this study a boosting regression tree model was built to study the factors and the influence path of housing prices from the microcosmic perspective. First, a classical hedonic price model was constructed to analyze and compare the significant effect factors on housing prices in the market segments of the three cities. Second, the gradient boosting regression tree method that is proposed in this paper was applied to the three markets in combination to analyze the influence paths and factors and the importance of the type of housing hedonic price. The influence paths of housing hedonic prices and decision tree rules are visualized. The significant housing features are effectively extracted. Finally, we present three main conclusions and several suggestions for policy makers to improve urban functions while stabilizing real estate prices.