Integrated Classifier And Its Applications In Personal Credit Evaluation

The application of ensemble classifier is studied by this paper in the field of personal credit assessment, and using the ensemble classifier of high classification accuracy and the ability to measure the importance of the characteristics of a variable feature, the traditional credit assessment model gets improvement and innovation. This paper has three innovation items. First, in terms of classification, using the ensemble classifiers, credit classification model are set up. Secondly, considering the particularity of credit evaluation, improved random forests and Boosting model are suggested, effectively reducing the first type of error. Thirdly, using random forests and Boosting model, respectively, two sets of personal credit scoring system are established based on ensemble classifier, combined with Chinese credit scoring models.This article discusses the common single classifier and ensemble classifier, the former includes discriminative analysis, K-Means algorithm, the smallest nearest neighbor algorithm, Logistic regression and decision tree methods, while the latter includes Bagging, Random Forests and Boosting model, respectively. Personal credit classification model are established using these models. Using the independent test set estimate the generalization error, and confirmed that the classification results in the comprehensive sense, the ensemble classifier is more excellent than the single one. Considering the specificity of credit assessment, this paper also proposed improved ensemble classifier.The personal credit scoring system, which based on the ensemble classifier, uses the characteristics of recognizing the importance of the variable to quantify the weight of each evaluation index, then to quantify the value for each variable using the risk measure method, finally a new scoring system is built. Empirical analysis shows that the new system can not only rate the customer’s credit level, but also identify and classify the credit risk. The effect on the classification is better than any other single classifier and ensemble classifier, and the new credit scoring system can well put into practice and is effective.