Study On Housing Price Forecasting In Beijing Based On ARIMA And PSO-BP Combined Models

With the continuous development of China’s economy,housing prices have gradually become an important object of concern.Through accurate prediction,it can provide a certain reference value of the house buyers and sellers.Therefore,the realization of house price prediction is a very meaningful research.This thesis takes the prediction of housing prices in Beijing as the research objective.First,we obtain the housing price data of Beijing from Anjuke website to January 2009 and September 2022 through web crawlers,and divide the processed data into training establish and test set.Housing price data before 2022 is used as the training set,the housing price data in 2022 is used as the test set.Secondly,Autoregressive Integrated Moving Average Model(ARIMA)and BP neural network model are taken into modeling analysis.It is found that ARIAM model fitting effect is good in the training set,but BP neural network model fitting effect is poor.However,ARIAM model prediction error is large in test set,and BP neural network model prediction error is reduced compared with the ARIMA model.In order to improve the fitting effect of the BP neural network model,the dynamic variable weight particle swarm optimization algorithm(PSO)is used to optimize the weight and threshold,and the PSO-BP neural network model is established,The experimental results show that the fitting effect of the PSO-BP neural network model on the training set is significantly improved compared with that of the BP neural network model,and the prediction error on the test set is decreased.In order to further improve the prediction accuracy of the model,the idea of combination model is introduced to construct ARIMA-PSO-BP series combination model and parallel combination model based on simple weighting method,dominance matrix method and residual reciprocal method.Finally,it is found that the fitting effect of the combined model on the training set is good,and the prediction results of ARIMA-PSO-BP combined model,ARIMA model,BP neural network model and PSO-BP neural network model are compared on the test set with the root mean square error,average absolute error and average absolute percentage error as evaluation indicators,and it is found that the prediction accuracy of the combined model is higher,Among them,the parallel combination model based on the reciprocal residual method has the best prediction effect and the lowest prediction error.