Application Research Of Machine Learning In Auto Finance Anti-fraud Model

China's auto industry has developed with the rapid development of the economy.China has become the world's number one auto consumer market and the world's largest consumer potential market.The growth of automobile production and sales has a significant driving effect on auto finance.China's auto finance penetration rate has increased from 13% five years ago to nearly 40%.With the increase of automobile financial penetration rate,the requirements for risk control ability are getting higher and higher,and related enterprises have begun to combine technology and automobile financial risk control.Due to the high amount of the car loan,the driving fraud agency falsified the information or swindled the creditless recorder to obtain the vehicle through the phased purchase of the car,and then transferred the car through illegal means.Therefore,the identification of anti-fraud risk for car buyers has become the key point of current risk control.This paper uses the sample data of Internet finance company users to explore the application of machine learning technology on the anti-fraud model,study the relevant characteristics of customers,and propose the establishment of anti-fraud rules.Establish a logistic regression model that is more recognized in industry and use it as a standard.Establish support vector machine,Adaboost,XGBoost anti-fraud model,introduce the main parameters of various models,and explore the optimal combination of parameters to optimize the model effect.Combined with the evaluation index of machine learning classification model,the performance of each model is compared.Finally,it is concluded that XGBoost has the best model effect in this study,and the performance of the model is greatly improved compared with logistic regression.Finally,the feature analysis and anti-fraud model are used to provide reference for the establishment of anti-fraud system,and the future anti-fraud system is expected to develop.