Study On The Application Of The Price Evaluation Model Of Second-hand Housing In Cities

The real estate industry is the pillar industry of our national economy,which is in forerunner position in economic market.The high and low of house price is closely related to the production and life of our people.In recent years,the change of domestic and foreign economic situation and the change of real estate prices present great uncertainty,the scale constraint of new housing and the instability of the price,so that many consumers turn their attention to the resold housing,the second-hand housing market has developed rapidly.As the main carrier of second-hand housing,the city has a large number of housing transactions.The traditional single price evaluation method has low efficiency and poor accuracy,which cannot meet the growing demand of second-hand housing market.Therefore,it is particularly important to grasp the law of urban second-hand house price change and seek an effective technology to evaluate the second-hand house price.In this paper,based on a review of literature and theories related to urban second-home price assessment,we use feature price models and machine learning methods to carry out innovative research in the following aspects.(1)First,we analyzed the key factors affecting the house prices from three aspects:location characteristics,building characteristics and neighborhood characteristics.Then the combination of web crawler and field research was used to obtain 2001 second-hand house data in Jinan city and quantify them.We also use random forest and Lasso regression methods for feature selection to construct a model containing the administrative area,which the neighborhood belongs,education support,building area.The index system of urban second-hand house price evaluation was constructed with 20 characteristic variables,such as the administrative area to which the district belongs,education support,building area,and external environment.(2)The linear model,double logit model,log-linear model and linear logit model were used for regression analysis.Based on the goodness-of-fit test and residual analysis,the loglinear regression model with the best fit was selected as the characteristic price model of urban second-hand houses for subsequent research.The results showed that the floor area,convenience of living,and distance from the city center had significant positive effects on house prices,the total number of households in the neighborhood and proximity to universities had significant negative effects on house prices,and the effects of characteristics such as house orientation,number of bedrooms,and floor area ratio were relatively insignificant.(3)Using random forest,lasso regression,XGBoost,Light GBM and KNN modeling methods,urban second-hand house price prediction models were constructed and empirical studies were conducted using Jinan second-hand house market data as a case study,respectively.The results of the goodness-of-fit test and evaluation error analysis show that the XGBoost model has the best evaluation effect,with a prediction accuracy of 92.24%.Finally,this paper puts forward some suggestions to promote the steady development of the second-hand housing market in Jinan from the perspectives of the government,real estate agents and consumers.