Research And Application Of Urban Real Estate Early Warning Based On Support Vector Machine

In recent years,with the rapid development of the country and the solid economic foundation,some new industries have emerged,which have provided many employment opportunities and accelerated the process of urbanization.At the same time,the real estate industry has flourished with the development of the economy and the acceleration of urbanization,and its irreplaceable role has become increasingly prominent in the national economy.The trend of the real estate market is closely related to the national macro-economy,and it will show some cyclical fluctuations,but if the fluctuation range is too large or the trend is too extreme,it will affect the development of the whole national economy.Therefore,it is necessary to observe and warn the real estate market so as to avoid its abnormal fluctuations.This paper consults the relevant literature at home and abroad,summarizes the theory and viewpoints for reference,and expounds the theoretical basis of the model,including the theory of real estate cycle fluctuation,the elements of early warning system and its functions.By comparing the advantages and disadvantages of the three commonly used early warning methods,a model early warning method with high adaptability for real estate early warning is selected,and a black-box model-support vector machine model early warning method,which has incomparable advantages in solving small samples,generalization ability and non-linearity,is selected.After the adaptability analysis of this method,the process of model building is described.Based on the early warning method of this model,Beijing,Wuhan and Kunming are selected as the representatives of different cities in China,as the empirical research object of this paper.The preliminary selection of real estate early warning indicators is based on the development of the local real estate market,combined with the principles of selecting early warning indicators and the availability of data.Through data collection,the data of early warning indicators from 2000 to 2017 are sorted out,and then the leading indicators are screened out from the 12 comprehensive indicators selected by the method of time difference correlation analysis.Based on the multi-class support vector machine,using the programming and operation of Matlab software and Libsvm toolbox,the screened warning indicators are processed as the input layer,and the classification representation of the real estate status of the early warning is taken as the output layer.Three classified learning machines are established to learn 17 training set data respectively,and the early warning model of the real estate market is constructed.Next,a detection set is input into the early warning model,and the classified output is compared with the actual situation.The results show that the model can predict the real estate market in the next year,which has a certain generalization ability.Finally,the collected warning indicators for 2018 are input into the model.The results show that the real estate market in Beijing,Wuhan and Kunming will run smoothly in 2019.At last,the results of the model are briefly analyzed,and policy suggestions are put forward from the aspects of land supply,credit scale,housing supply structure,restriction of speculation and strengthening the transparency of real estate information,so as to promote the healthy operation of the real estate market in China.