| Real estate is a pillar industry in China.Since the 90 s of last century,China's real estate industry has maintained a sustained rapid growth,which has provided the impetus for the development of China's economy and the development of GDP.But the high prices also greatly limit the development of China's economy and improve the cost of our enterprises.At the same time,the high price of house prices has also brought great pressure to the general public.Based on this,the prediction of house price becomes very important.It is noted that housing types and real estate market activity vary widely,and house price index excludes the influence of housing quality,building structure,geographical location and sales structure,only considering the price fluctuation caused by supply and demand and cost fluctuation,so this paper chooses price index as a reference index instead of house price.At the same time,considering the possible factors that affect the housing price index,there are CPI,money supply,Shanghai Composite Index and exchange rate.Because the relationship between the various factors is not a simple linear relationship,considering the complexity of the relationship between variables,the traditional model is difficult to measure,so this paper makes a breakthrough to consider the above indexes as input variables and use theBP neural network to model.Then,the prediction results are compared with the actual results,and then the number of variable input variables is considered to optimize the model,and the new prediction results are obtained.In addition,we set up a time series ARIMA model for housing price index.Finally,we compare the results of several models,draw the conclusion,and interpret the practical significance of the prediction results.By comparing the error between the neural network model and the prediction result of the time series model,we can find that the error of the prediction results of the four input variables neural network model is the least,while the ARIMA model has the greatest error.It shows that the neural network model is better than the one time series model in fitting the housing price index. |
