Research On Forecasting Method Of Urban Real Estate Price Index In China

With the steady development of the national economy and the improvement of the people's living,people pay more and more attention to housing prices.Not only that,housing prices have always been a topic of common concern to scholars at home and abroad,and have always been a hot research field.Understanding and grasping the future trend of housing prices is crucial,not only to help home buyers to configure their properties,but also to help real estate developers make development decisions and assist government departments in implementing policies.In order to solve the problem of precision loss caused by the fixed parameter setting of traditional prediction methods,this paper introduces the dynamic model method,namely dynamic model averaging(DMA)and dynamic model selection(DMS).Based on this,the corresponding forecast research on urban real estate price index in China is carried out.The dynamic model method allows the predictor settings contained in the equation and the equations contained in the model to change over time.Experiments show that this time-varying mechanism makes the dynamic model method have a good effect on the prediction of house prices.In addition,this paper examines EW,TVP-AR,BMA and other models as experimental comparisons to further illustrate the superiority of dynamic model prediction.The main work of this paper is as follows:1)Research on mining of key predictorsSince the data set of housing prices has not been publicly researched in China,we have studied the factors affecting housing prices after referring to the previous studies,and then created a data set of six different levels of urban housing prices based on this,and finally applied the machine learning algorithm performs the calculation of the importance of variables to complete the mining work that affects the key predictors of house prices.2)Research on house price forecasting model based on DMSThe traditional house price forecasting models can meet the forecasting requirements to a certain extent,but the accuracy of the models are not satisfactory.The main reason is that the fixed model parameters and variable settings of the traditional methods will lead to the limited prediction ability of the models.Therefore,this article will use a dynamic model with a time-varying mechanism that allows the number of model variables and variable coefficients to change over time creatively.The experimental results show that the dynamic model method has smaller prediction error than the comparison models.3)Research on optimization of price forecasting model by network search indexesThe Internet search index of the big data environment background has a very good effect on the prediction of some macroeconomic variables in the future.Therefore,this paper optimizes the predictive performance of the model by constructing a special data crawler framework and obtaining Baidu search index with strong keywords,and then,Baidu search variables and economic variables with different keywords are used as variables to optimize the prediction performance of the model.