Housing Price Index Prediction Based On Web Searching Keywords

The popularity of the Internet influence the people's lifestyle and consumption habits unconsciously.People habit to make decisions especially consumer decisions through the search engine because the Internet contains massive network data.The search engine is the bridge between consumers and information and the the search keywords is the key of looking information.So,the search keywords represent the focus of consumers to a certain extent and search volume represents the degree of concern.It can be concluded that the network search keywords data mapped the market subject of concern,revealing the trend of market behavior that can provide the micro data for macroeconomic issues.Various scholars pay more and more attention to housing prices,especially in large and medium city residential price because the residential sales industry is the key impetus to the rapid development of the nation economy and society.This paper research the relationship between Xi'an housing price and network search keywords data.This paper analysis on the main factors influencing the housing price from macro and micro perspective based on the equilibrium price theory.From the qualitative point paper expound the relationship between the new commercial housing sales price and the network search keywords data.Paper select 64 keywords as the primary thesaurus from the point of view of the factors of residential prices using the method of seed keywords and expanded keywords.Paper use the seed keywords to do gray correlation degree method.Then,using Spearman correlation analysis screen high correlation search keywords and using time difference correlation analysis screen peer keywords.Using principal component analysis method carry on the synthesis of keywords forming four comprehensive indexes.Finally,building the Neural Network model,the Random Forest model and the Support Vector Machine model.Paper analysis the predicting performance of these three model and find that the optimal model is the Random Forest model.Using GBDT analyze the prediction results of three model again to get the synthesize residential sales price index.The main conclusions are as follow:(1)web search keywords and Xi'an housing price have a relevance;(2)the prediction results of GBDT model is best and the fitting degree is 0.995;(3)using network search keywords data can be compared to the Nation Bureau of statistics in advance for half a month to count the residential price index..