Application Of Data Mining Model In Personal Credit Risk Control

In recent years,with the rapid development of China's economy and the improvement of residents' income level,more and more people begin to get used to consuming in advance,which promotes the rapid growth of China's personal credit business market.However,in the process of development,many problems gradually emerge.First,the traditional bank review mechanism is too tight,which will not only greatly increase the bank's cost of work,indirectly increase the loan interest rate,but also easily make it miss quality customers,affecting the bank's earnings;Second,the review mechanism of the emerging online loan platform is too loose,making the two sides in the information asymmetry position,resulting in many bad events such as customer credit default,platform running away with money,which has a very bad impact on the development of China's credit market,and at the same time poses a severe challenge to the risk control of credit business.The root cause of these two problems is that the lender cannot make an accurate judgment on the customer's credit level,so that he cannot make a correct lending decision.Data mining is a means of analyzing massive data.Through the training process of various models,we can find the hidden and imperceptible association relations and causal rules in the data.In the field of credit risk control,after years of market development,the industry has accumulated a lot of customer loan credit data,the data itself not only contains great value can be analyzed,and as a result of the existence of the above credit market problems,analysis of these data and find out the level of customer credit rules become necessary and urgent.Through the application of data mining model,people can find the relationship between various variables in customers' historical loan data,thus providing help for judging customers' credit level,further predicting customers' default risk,and providing strong support for lending decisions.This paper first introduces the relevant concepts and theoretical basis of data mining to prepare for the subsequent model training and prediction.Then,taking the LendingClub data in the field of personal credit risk control as an example,various means such as variable selection,data transformation,data dimensionality reduction and class imbalance processing were adopted to preprocess the data.After that,F1 measurement is taken as the model evaluation index,and various data mining models including K nearest neighbor,logistic regression,decision tree andsupport vector machine are applied to the data,and model optimization is carried out on the basis of these traditional individual models to improve the prediction effect of the individual model.On this basis,the Stacking in the ensemble model is applied to the data.This model not only synthesizes the predicted value of the previous individual model,but also solves the combined strategy of the individual model by training the secondary model,which greatly improves the prediction ability of the model.Finally,the trained Stacking model is used to predict whether the customer will default,and the corresponding lending decisions are made according to the prediction results.It is found that the default rate of the customer is lower than that when the data mining model is not applied,which plays a role in risk control.