Research On Xi’an Real Estate Price Forecast Based On Improved LSTM Model

The good development of the real estate economy plays an important role in the stability of the national economy.With the continuous upgrading of the market environment,the real estate industry has entered a new high-quality stage and is advancing in a more benign direction.The fluctuation of real estate prices is also a livelihood issue that has become a great concern.Therefore,this thesis takes Xi’an city as an example and conducts a research on real estate price prediction based on an improved LSTM model.The main content is divided into three parts.First of all,this thesis analyzes the factors affecting the real estate price,and expounds the correlation mechanism between the network search data and the real estate price fluctuation.On this basis,the thesis determines the initial key words by taking technical words as the main part,obtains the expanded key words base by using the related word meaning expansion methods,and screens them by using Spearman correlation analysis and time difference correlation analysis,so as to ensure that the selected key words are important and advanced,and builds the screened key words into the network search data index system.Secondly,in order to eliminate the noise disturbance in the network search data,we combine the complete ensemble empirical mode decomposition with adaptive noise with the improved wavelet thresholding to decompose and reduce the noise of each index,so as to improve the quality of the data,and test the ADF stationarity of the index after noise reduction to ensure the stability of the data and avoid the occurrence of "pseudo-regression".Finally,based on the analysis of the principles of the benchmark models,PSO is used to optimize the two-way long-term and short-term memory neural network model.Combining with the forecast results,some suggestions and measures are put forward from the perspectives of consumers,investors,real estate developers and the government.Based on the research,there is a high correlation between the network search data and the real estate price,and it has a certain lag in advance of the price fluctuation.