Real Estate Price Index Forecast Based On Web Search Keywords ——A Case Study Of Guangzhou City

In the wake of Internet era,the consumption custom and the way of life of the public change accordingly.People use the search engine as a medium to obtain information through the Internet,at the same time,the Internet also stores a large number of user behavior information.The behavioral information contains the users' behavioral intention and the search volume can reflect the public's attention timely in a certain degree,which also affects the market fluctuations.The real estate is the pillar industry of China and it is paid high attention by scholars.However,the publish of the real estate price index with low timeliness and long-time consumption.Therefore,this thesis will forecast and research the real estate price index through keywords data of Internet search that could reflects the behavioral information of the public timely.This thesis,taking the price index of new-built commercial residence as the subject,and based on the supply and demand theory,expectation theory and behavior theory of real estate market,analyzes the relationship between internet search keywords and the price index of new-built commercial residence from the qualitative perspective.Then extract the keywords of large amount of news and forum text about Guangzhou real estate by NLPIR semantic system and obtain the core keywords of this thesis through extending sieving,expanding to a lexicon containing 48 internet search keywords.After that,through Spearman correlation and time difference correlation analysis,12 antecedent keywords are sieved.After principal component analysis,4 comprehensive indicators of Guangzhou internet search keywords are gained.Finally,the random forest and the support vector regression model are established by using the comprehensive indicators,and the parameters of the support vector regression model are optimized by the particle swarm optimization algorithm,and the three models are evaluated.The results show that the forecasting effect of the stochastic forest is the best,and the fitting degree reaches to 0.996.The following three conclusions are drawn :(1)There is a strong correlation between Internet search keywords and the price index of newly-built commercial housing in Guangzhou.(2)It has a good effect to predict the price index of newly-built commercial housing in Guangzhou by constructing a model based on the comprehensive index of Baidu index in Internet search.(3)the prediction results of the price index of newly-built commercial housing in Guangzhou by using Internet search keywords were published half a month earlier than that of the National Bureau of Statistics.