Risk Assessment In Real Estate Project Based On Random Forests

China’s real estate industry has completed the transition from the “golden age” to the “silver age”,the real estate project risk problem has become increasingly prominent.This study aims to study the application of random forests algorithm in real estate project risk assessment,and expects the research results could improve the efficiency and effectiveness of real estate project risk assessment,to help real estate enterprises to find and avoid risks in project decision-making stage,to reduce human resource costs and to improve business performance.This study uses the core tools-random forests algorithm,supplemented by literature research,interviews,qualitative analysis,correlation analysis and other methods.Based on the perspective of J Real Estate Corporation,this paper studies the potential risks of 48 local residential land.In the research process,the risk evaluation index system of real estate project including 7 dimensions and 55 indicators was established.On the basis of above,the study first excluded 28 indicators that had no difference in the 48 studied land,and then excluded nine indicators that were highly relevant to other indicators based on the Spearman correlation analysis.In this study,the evaluation results of each qualitative index and the risk level of each sample were obtained through the interview of executives of J Corp.,and the data of the quantitative parameters,satisfied the data requirements of the random forests algorithm.Moreover,an analysis on the random forests model is carried out.It is found that the model has an accuracy of 95.6%,which can effectively identify the risk of real estate projects.And 5 most important indicators are also obtained: project location,education resources,urban disposable income,urbanization rate,regional attractiveness.Finally,I conclude that the accuracy of RF is highest to the data in this study after this research compares RF with kNN and SVM by 10-fold crossvalidation.In this study,the evaluation system is broad and the index is abundant,which have a certain reference meaning to the future real estate risk related evaluation.This study also takes the lead in applying the random forests algorithm to the real estate project risk assessment field,and provides the future similar assessment a new idea.