Car Dealer Credit Risk Analysis Based On Integrated Machine Learning Algorithm

Recently the Automotive Industry has been in a depression.The auto dealers with single franchise authorization can only sell vehicles of one brand,which makes huge damage to their profitability.Banks and other financial institutions tightened their credit policy towards the dealers,which leads to the liquidity issue of the dealers.Once the dealer's funding support is broken,they will default and cause credit loss to the financial institutions.Short-term credit risk of the dealers is increased.How to conduct credit review and regular risk monitor rapidly and effectively towards massive of dealers becomes the priority issue for the Auto finance company.The auto finance company I work for has an internal credit risk score card and will monthly rank the dealers upon their financial and operating performance.Credit analyst will take relevant actions according to the final score.However,the credit model has been launched for years with all variables provided by foreign experts and imperfect procedures to review the accuracy of the model,which leads to the problem from obsolescence,variable explanation and forecast accuracy.In order to make decision towards dealers more effectively and conduct regular risk monitor,this paper wants to establish a new score card model based on previous model and working experience in auto finance company.This paper will go through five steps.First,it will introduce the background of auto finance,through literature review to learn score card model.Second,it will introduce the theory of machine learning.Third,it will expand variables and conduct comprehensive analysis.Forth,it will establish Logistic model and XGBOOST model with SAS and Python.Comparison will be made from explanation ability and forecast accuracy considering the imbalance of data category.Then build a new scorecard with WOE exchange.The final part is conclusion and going forward monitoring.