基于RGL-Stacking模型的存量住房价格预测研究

t present, China has entered the era of stock housing, and the asymmetric transaction information of stock housing leads to the listing price being generally higher than the transaction price, which causes losses to the participants in the transaction and is not conducive to government departments to formulate policies. Therefore, it is particularly important to accurately predict its price. At present, machine learning is widely used to predict housing prices, but most of the models used are single models, and single models are prone to over-fitting and lack of expressive ability. In order to overcome these shortcomings, based on the stock housing transaction data in Nanjing, this paper selects 11 factors that have the greatest impact on the transaction price by using correlation analysis, and establishes three single models, namely random forest model, XGBoost model and neural network model, and RGL-Stacking as a fusion model. The results show that the mean square error MSE and the determination coefficient R are used as the judgment basis. The determination coefficient R of RGL-Stacking model is 0.864, and the average square error MSE is 2772, which proves that RGL-Stacking fusion model has more advantages in forecasting housing prices.