Empirical Research On Sales Forecasts Of Second-hand House In Wuhan Based On XGBoost Algorithm

With the rapid development of big data and internet technology,there are more and more datas produced in the real estate industry,it is vital important to get the house price forecasting model from the complex datas.The purpose of this paper is to find out the important characteristics from the large number of datas of second-hand house in Wuhan,and to set up the price forecasting model.Based on related house price theories,there are three types of characteristics that impact house price,including location,structure and neighborhood.Using web spider,we can fetch those datas specifically.Combining with the previous results,we choose XGBoost algorithm.It is quite efficient in classification and regression with pretty high efficiency.This model can adapt balanced and imbalanced data,at the same time it is not easy to over-fit,and has good generalization capacity,been used widely.By applying grid-search,we get the best parameters for the model.Finally we compare the forecasting results from XGBoost with that from LASSO,it shows that XGBoost has a significant advantage.