Research On The Multicollinearity In Liability Analysis

The problem of credit assessment of online loan users is of great significance to online loan companies.This paper tries to predict the future liabilities,to provide intelligence about credit assessment.Linear regression method is one of the most classical statistical methods.This paper will use linear regression model to predict users' future liabilities and discuss the advantages and disadvantages of several methods.In multiple regression,multicollinearity may lead to many problem.In order to reduce or eliminate the impact of multicollinearity,this paper mainly introduces the traditional biased estimation method and machine learning algorithm.The research contents can be divided into the following parts:Firstly,the influence of least square method and multicollinearity on it are briefly reviewed,and the definition,effect and feasibility of liability analysis are introduced.Then,the biased estimation method and machine learning method are introduced in two parts: Liu estimation,principal component estimation and Lasso estimation,support vector regression and XGBoost model.The related theories,implementation processes and the principles of improving multicollinearity of these methods are discussed respectively.Finally,the data are obtained for case analysis,and the data of May 2018 is used to predict the users' liabilities in August.The prediction results are obtained and the applicability and advantages and disadvantages of several methods in this case are compared.Finally,a conclusion is drawn.Aiming at the problem of liability analysis,this paper compares the prediction effect of several linear regression algorithms,and provides examples and basis for future application.