Application Of The Weighted Quantile Regression With Adaptive LASSO Method In Overseas Property Valuation

With a boom in overseas real estate investment,the demand for house property valuation is expanding.However,the valuation methods commonly used have certain defects that the hedonic price model is not robust and the machine learning method is lack of explanation capacity.To this end,this paper proposes the weighted quantile regression with adaptive LASSO method(WQR-LASSO)to evaluate overseas real estate.The WQR-LASSO method proposed in this paper can achieve robust parameter estimation and variable selection simultaneously.And with no assumption about the distribution of error,the method can describe the whole picture of the conditional distribution of housing prices.It not only develops quantile regression theory,but also improves existing property valuation methods and provides technical support for investors to evaluate overseas house property.The simulation results show that WQR-LASSO performs better on finite samples and has higher robustness than adaptive LASSO and LAD-LASSO,whose correctly fitted rate is 100% and MSE is smaller than 1.In the empirical analysis,WQR-LASSO is used to evaluate properties in the Greater Boston,which provides more accurate valuation compared to the hedonic price model and popular random forest.WQR-LASSO of 0.75 quantile selects 7 significant variables with 0.13 MSE.Using its estimator to explore the factors of different housing prices in the Greater Boston,the following conclusions are drawn.(1)The heterogeneity of low and high price housing is stronger,so the factors are more various.Whereas,housing of middle price has certain homogeneity,so its factors are fewer,which is more obvious with the rise of house price.(2)Education is the most important factor of housing prices,and room,location and crime take the second place.(3)Different housing prices have their own factors.