Research On Credit Rating Method Based On Rule-Classifier

With the rapid development of information technology and the global economy,the operating environment of financial institutions has undergone tremendous changes and is full of suddenness,variability and complexity.The control of credit risk is the operating core of financial institutions.The ability to control credit risks also reflects the core competitiveness of financial institutions.At present,the risk management level of China's financial institutions is still not high.Taking into account the reasons in this regard,the development of a rating system and model that can accurately integrate customer personal information and risk characteristics,and have better predictive analysis capabilities,is effective in improving finance.The inevitable demand for the organization's risk management capabilities.In summary,the R&D financial user credit rating system is used to complete the rating work of financial users,and while promoting the development of domestic financial users,it has also effectively improved the rating system for financial users and effectively managed the credit risk of financial users..This article comprehensively considers the current credit risk situation faced by financial institutions,systematically analyzes the existing mainstream financial customer credit rating methods,including traditional credit scorecard models and credit rating models based on logistic regression.A financial user credit rating system based on Ripper rule classifier is proposed.Ripper algorithm is a classification algorithm based on rule induction.The algorithm itself has high processing efficiency,and it can still run efficiently on data containing many noises,and it has high anti-noise ability.And because Ripper's algorithm has good scalability and regularization,it solves the disadvantages of poor targeting and difficult interpretation of the scoring model,which makes it easier for policymakers to understand their judgment criteria.For multidimensional feature attributes,it is different from the previous single model.The feature selection algorithm adopts a feature selection algorithm based on Ripper.Combined with the features of the classifier itself,it is more intelligent and controllable to gradually filter redundant attributes,and the IV value algorithm can achieve more optimized and more accurate features.Select method.Finally,the three models were compared and analyzed using the auction loan company data.Experiments show that compared with traditional scoring cards and logistic regression-based credit rating systems,the credit rating system based on Ripper rule classifier proposed in this paper is compared with the logistic regression-based credit rating system.The classification accuracy for users can reach 91.27%,which is higher than traditional score cards and logistic regression methods.