Application Of Data Mining Technology In Commercial Bank Personal Credit Risk Assessment

In recent years,with the transformation of China's economic structure,the level of resident consumption has gradually increased,and various types of consumer credit business of commercial banks have surged.This has made bank risk management face enormous challenges.The traditional risk assessment method mainly relies on the market experience of credit officers.The risk assessment results are subject to subjective factors of individuals,and in the face of increasing data volume and business volume,the traditional assessment methods are inefficient,the business cycle is long,and the accuracy rate is low.It is also difficult to guarantee.Therefore,the traditional assessment method can no longer meet the development of the current commercial bank credit business,and commercial banks urgently need to establish a new type of credit risk assessment model with high assessment accuracy.This paper takes the practical application of commercial bank's personal credit business as the background,takes the German bank's credit data collection donated by Hamburg University as the research object,and studies the commercial bank's personal credit risk assessment model.The main work is as follows:Firstly,the original data is preprocessed,including removing missing values and outliers,merging of variable features,data transformation,data standardization,etc.,and using the processed data for descriptive statistical analysis to construct a simple credit client portrait.Secondly,three single algorithmic personal credit assessment models are established,which are logistic regression model,decision tree model and neural network model.The performance of single algorithm model is optimized by using parameter adjustment and introducing cost matrix.Finally,the prediction results of each model show that the overall prediction accuracy of the three models on the test set is above 65%,and the prediction effect is good.Then,in order to further improve the prediction accuracy of the model,based on the single algorithm model,this paper also established an ensemble algorithm model and a combinatorial optimization model.The ensemble model selects widely applied boosting integration model and random forest integrated model.Compared with the single algorithm model,the prediction accuracy of the two ensemble algorithm models is significantly improved.In addition,the variables were sorted by using the size of the mean decrease accuracy(MDA)of each variable in the random forest algorithm,and 15 variables that had a great influence on the model results were selected.The combination optimization model is based on the neural network single algorithm model,aiming at the problem of the poor stability of the model,the genetic algorithm is used to optimize the model,and the random forest algorithm is used to filter the variables.It is found that the combination optimization model has smaller prediction standard deviation for training data and test data,and the stability of the model is obviously improved,and the prediction accuracy of the model to the test set is 83.5%,and the prediction effect is the best.Finally,combined with the practical application background of commercial bank personal credit assessment,the six models established were compared and analyzed.The results show that the combined optimization model has good performance in model prediction accuracy,model interpretation and stability,and the model performance is better than other models.Therefore,the combinatorial optimization model has certain reference and application value for commercial banks to establish an automated credit evaluation system.