Research On Housing Pricing Model Based On Machine Learning

As the number of social commercial housing increases,the total number of residential transactions is increasing.In the residential transaction scenario,the buyer and the seller may be said to be in the position of both sides of the game.It is of practical significance to study a fair and scientific housing pricing method,which will greatly facilitate the residential transaction,and achieve a win-win situation between the buyer and the seller.In this paper,a machine learning model of housing pricing is built based on the internal characteristics of housing and its sales price.The research of this paper mainly focuses on the following three aspects:(1)Data analysis and data preprocessing.Firstly,exploratory data analysis is carried out,including data distribution and correlation analysis.In order to achieve a better effect of the model,characteristics are preprocessed,including missing value processing,outlier processing,data transformation.(2)Using machine learning algorithms to construct housing pricing model.LASSO regression,support vector machine,random forest,GBDT and XGBoost were respectively used to construct the housing price regression model.The parameters of each model were adjusted to achieve the best prediction effect of each model.The prediction effect of each model was evaluated and compared.(3)Using averaging and Stacking to improve the pricing model.The Avg_Stacking ensemble method is proposed to improve the former machine learning models.The experimental results show that the prediction error of the Avg_Stacking ensemble model is about 24%less than the random forest,which can predict the housing price more accurately and provide a scientific and effective reference for reasonably estimating the housing value.