Machine Learning Approaches to House Price Determinants

Abstract

This study takes housing prices in Jinan as an example. Data were collected from official government websites and official housing price platforms. The Hadoop platform was used for distributed storage and preprocessing. Linear regression and recurrent neural network (RNN) models were constructed to compare and analyze the influence of various factors on housing prices. The results show that factors such as interior area, location distance, and regional planning have significant effects on housing prices. Among these, the marginal contribution of interior area is the highest, and communities located 600–900 meters from a subway station have the optimal prices. The RNN model achieves better prediction accuracy than linear regression and effectively captures market dynamics, nonlinearity, and time-dependent characteristics. The comprehensive ranking of influencing factors is: interior area > regional location > surrounding planning development > housing condition > shared area. This study indicates that interior area and regional location are the main factors affecting housing price fluctuations. The findings can provide data references for homebuyers, developers, and policymakers.