Housing Price Forecast Analysis Based On Web Search Data

With the development of the Internet,search engines have become an important channel for people to acquire knowledge.The "search" traces left by netizens on search engines form network search data,and the search data accumulated by search engines provides important information resources for people's research.Based on this,through the data mining of the network search data,it is possible to effectively capture the purchase intention of the consumer,the investor's attention before the investment,and then predict the price level of the commodity transaction link.Real estate has always been the focus of the whole society.The fluctuation of housing prices is related to the national economy and the people's livelihood.The effective forecast of housing prices is conducive to the reasonable investment of residents,and it is convenient for the government to propose macro-control policies and maintain social and economic stability.Located in the southwestern economic center,Chongqing has a high population concentration and is a core city of national concern.Its housing prices directly affect economic development and people's happiness.This article will use the monthly Chongqing new residential price index to measure the monthly housing prices in Chongqing,using the online search index to predict the new residential price index in Chongqing.The traditional method of forecasting house price index is to empirically screen macroeconomic indicators to establish a predictive model.Since the National Bureau of Statistics released a macroeconomic indicator only one year,the macroeconomic indicators are obviously lagging behind,and the empirical screening variables are subjective.The network search data is time-sensitive,avoids the lag of traditional methods,and uses data mining to filter variables to predict house prices,overcoming the subjective disadvantages of traditional methods.This article conducts research in five steps.First,use python to crawl the relevant news texts about "Chongqing House Price" on Sina.com,360 News Network,Phoenix News and other news websites as the initial corpus.Second,use the NLPIR system to extract the initial corpus based on the initial corpus.Key words,after the de-emphasis step,the initial factors that are closely related to house prices are initially obtained.Thirdly,22 final keywords are selected by the time difference correlation coefficient method and the random forest average precision reduction method.Fourth,through Baidu The index platform obtains the Baidu index corresponding to the final keyword,and obtains the monthly Baidu index after processing.Fifth,the monthly Baidu index of the final keyword and the Chongqing new residential price index released by the Bureau of Statistics are used for modeling,and the modeling process is adopted.Traditional econometric model least squares regression and machine learning regression models.The main conclusions of this study are as follows:(1)According to the time difference correlation analysis method,the keywords are most closely related to the newly-built residential price index in Chongqing,which is the “public rental housing”.It can be seen that with the rise of housing prices,residents' residential pressure Increased,public rental housing is getting more and more attention.(2)After the random forest average precision reduction study,the key word for predicting Chongqing's new residential price index is “love”,the keyword “marriage” is followed,and the housing in China is just the need of marriage and love.An important driving force for sales.(3)Compared with the traditional least squares regression,the gradient elevation regression in machine learning is the most stable and accurate in predicting the new residential price index in Chongqing.