The Price Factor Analysis Of Second-hand House Based On The Method Of Data Mining And Lasso

With the continuous development of economy and the rapid advance of urbanization,prompted the rapid development of the Real Estate Market in China,at the same time real estate related economic activity is becoming more and more frequent.Due to less and less land to cities for development,and second-hand housing transactions have become more active,people's demand for real estate information and valuation is growing.Both from the perspective of market participants,and from the perspective of Countries began to levy,second-hand housing prices accurately measure is a timeless topic.Fast accurate evaluate of real estate,not only can provide the seller with appropriate evaluation price,at the same time provide scientific price forecast to the buyer,and can guarantee both parties more efficiently promote the business.On the real estate price appraisal in China are generally adopt three traditional assessment methods:the market comparison method,cost method and income method.Market method in the evaluation are mostly rely on the experience of the evaluator,easily affected by subjective evaluator.In recent years,in order to improve housing assessment methods,many scholars began to introduce statistical modeling method to the real estate price appraisal,and obtained the very good effect.This paper will draw lessons from foreign experience of real estate appraisal,use the method of data mining to set up the second-hand housing price evaluation model.Collected more than 30000 second-hand housing information from Lianjia sites in Beijing by means of web spider,including architectural features,geographical features and such so on 38 of the key factors affect the price of second-hand housing.First of all,used feature selection and Lasso regression two kinds of method to carries on the preliminary feature selection,to get rid of the less influence factor from the model,reducing the complexity of the model,finally selected 30 variables to the next step of model.Second-hand housing price evaluation model mainly constructed the traditional Lasso model and four kinds of data mining model: regression tree,Boosting,Bagging and random forest.Compared the 5 kinds of prediction model by the method of Fifty percent of cross validation,the results show that the random forest had the minimum error and had better fitting effect.At last,for the model of random forest make parameter adjustment and the model optimization,and set of test data tomake predictions,the test model fitting effect is good,the accuracy of the predicted results is very higher.Indicates that the second-hand housing assessment model based on random forest methods is a worthy of application and popularization of the method in real estate appraisal.From the perspective of the variable importance of random forest,urban area,building area,property costs,distance from the subway to walk the short distance and so on.That is the main factors influencing the secondary housing prices in Beijing.