Research And Implementation Of Bank’s Credit Risk Monitoring Model Based On Deep Neural Network

The appearance of credit risk will wreak havoc on the development of a country’s financial and economic. However, the development of credit risk management methods and technology is very slow in China, most of the state-owned commercial Banks is still in the primary stage of simply using foreign control model and basic case confirmed, lacking the ability of financial risk forecast and global control in advance. The current global economic situation is still not optimistic and very undulatory, thus banks need to rely on accurate predictions to control risk. According to the characteristics of commercial bank credit risk in our country, establishing a credit risk monitoring model is needed.The development of artificial intelligence technology and the rise of large-scale bank data provide a new direction to establish a more effective credit risk monitoring model. Deep neural network(DNN) is one of the most effective and the most widely used artificial intelligence technology, and it has the ability of learning excellent feature and fitting large-scale data. Considering the actual management mode and management environment of commercial banks in China, on the basis of referring the experience of foreign commercial bank credit risk management, we presents a bank s credit risk monitoring model based on deep neural network. The contributions of this work can be summarised as follows:1. Through the in-depth analysis of the influencing factors of credit risk, strictly following the principle of indicator construction, and according to the characteristics of the bank loan business, we build a relatively comprehensive monitoring indicator system of credit risk. The indicator system contains both the important financial income indicator, also includes cash flow, corporate growth, scale, and other related indicators, can comprehensively reflect the risk status of loan enterprise. Then the more important and more discriminative indicator can be futher obtained by indicator filter method.2. Aiming at the shortcomings of the traditional methods, credit risk assessment using deep neural network is proposed, and we design the suitable model structure and parameters for the actual risk assessment. In this paper, deep neural network methods, the latest artificial intelligence technology, the depth of the neural network method, is applied to the field of credit risk monitoring, which widen the methods and ideas of bank’s risk monitoring and management.3. Exampled by the bank credit risk status of a certain branch of Agricultural Development Bank of China as a case study, combined with the common problems of bank in our country, we propose a new model of the risk monitoring. The proposed method not only uses the traditional mathematical methods such as significance test, factor analysis, but also uses the latest deep neural network technology. Experimenttal results on the actual data demonstrate that our method has the advantage of high risk identification accuracy, and can be used in the actual risk monitoring compared with the traditional k-neighbor method and the logistic regression method.