Design And Implementation Of Credit Risk Assessment System Based On Flink

Since the concept of Inclusive Finance was formally introduced into China in 2006,more and more attention has been paid to the development of Inclusive Finance in China.In recent years,our government has also promoted the development of Inclusive Finance Strategy,gradually implemented various preferential measures,and strive to establish a convenient and equal financial service system for all social strata and groups.Although the strategy of Inclusive Finance in China has been promoted very rapidly,some problems have been exposed in the development of Inclusive Finance due to the inherent profit tendency of financial capital.For the key groups of Inclusive Finance,most of the relevant credit systems are missing,and the construction is more difficult;Furthermore,in the past few decades,the main customers of China's commercial banks are large and medium-sized enterprises,the business development and start-up of inclusive financial services for small and micro enterprises,farmers,low-income groups in cities and towns are relatively late,which makes the previous credit risk control rules not applicable to the current form of credit business.Under the above background,this paper analyzes the defects and deficiencies of the classical ensemble learning model in the credit risk assessment scenario under the background of Inclusive Finance,and then introduces the information entropy theory and cost matrix into the stacking integration model,proposed the EC-Stacking model suitable for the credit risk assessment scenario of Inclusive Finance,and the calculation method of cost matrix in the assessment process.Then,this paper uses about 900 thousand credit data of Lending Club in 2016 and 2017 as samples to calculate and compare the differences with other models in the relevant performance indicators.The results show that EC-Stacking model can greatly improve the discrimination performance of default samples by sacrificing the discrimination performance of normal samples within an acceptable range,and can effectively reduce the credit risk.It verifies the effectiveness of EC-Stacking model in credit risk assessment scenario.Then,aiming at the characteristics of "Large Volume" and "Low Delay" of Inclusive Finance,based on Apache Flink real-time computing engine and EC-Stacking model,this paper designs and builds a big data service architecture that can assess the credit default risk in real time,and designs and implements a credit risk assessment system according to the process of software engineering.In this paper,the functional requirements and non-functional requirements of the system are analyzed in detail,then the overall architecture of the system is designed and divided into four functional modules:data management,model management,risk assessment and user management,and each functional module is designed and implemented in detail.Finally,the credit risk assessment system is tested from two aspects of function and performance.It confirms that the system has complete functions,great performance and operability,and meet the functional and performance requirements of credit risk assessment system in the context of Inclusive Finance.