A Research On The Forecasting Effect Of Commercial Housing Transaction Volume Based On Internet Search Data

As a pillar industry in our country,the real estate industry has always been highly concerned by the public and the transaction situation in the real estate market has been given a lot of attention.However,the National Bureau of Statistics generally releases statistical data on real estate market transaction in the middle of each month,which is obviously lagged behind and cannot meet the needs of governments for making policies in a timely manner.Therefore,the lack of real-time official data has become a key issue restricting the rational decision-making of government or real estate market participants.Fortunately,the rapid development of Internet technology has provided new method for solving this problem.Today,when people want to buy houses,their expectations and behaviors will be shown in the real estate market and on the Internet.The behaviors are reflected in the real estate market by changes in the transaction price and volume of commercial housing,while the expectation is reflected in the Internet by changes of some indexes,such as search volume of relevant keywords.Therefore,the volume of internet search for real estate market information may be correlated with the commercial house transaction volume.Based on the above logic,this paper researches on the relation between the market transaction and Internet search indexes using Baidu search keyword data in the cities of Beijing,Shanghai,Hangzhou,Wuhan and Chengdu.It firstly comprehensively determines the keywords related to the volume of commercial housing transaction,and selects the top ten keywords by the methods of random forest and time difference analysis respectively.It composes some Baidu search indexes by principal component analysis.Then this paper constructs an autoregressive model and some models with Baidu search indexes for volume of commercial housing transactions respectively,and compares the forecasting effects of the models through short-term prediction and long-term prediction.By analysis of the research results,this paper draws the following final conclusions:(1)In general,Baidu search data can improve the forecasting effect of commercial housing transactions;(2)The Baidu search index data is more suitable for short-term prediction of commercial housing transactions,which can improve the timeliness of prediction;(3)Random forest analysis is better than time difference analysis to select keywords;(4)Home-buyers' focuses on real-estate-related information are concentrated and are greatly affected by macro policies.