Research On Method Of Customer Credit Risk Warning Based On Massive Financial Transaction Data

Under the circumstance of global economic, the degree of the financial industry’s openness and market is constantly increasing. Credit risk is still the most important risk, which will not only affect the bank but also have an influence on the customer. Research on how to use the bank’s existing data-massive customer financial transaction data has the important theoretical significance and application value for quantitative credit risk management and early warning.This thesis designs and achieves Customer Financial Transaction Data Analysis Platform based researches on the core business data in a commercial bank, which contains massive and distributed heterogeneous financial transaction data. Thesis work and contributions of this article are as follows:1. Make a further research on the more business and management approach related to bank loans, on this basis, and focus on the commercial bank borrower’s financial transactions.2. Optimize customer transactions based on commercial bank credit, as well as predict customer repayment probability of default model. By training SVM-classifiers in the decision-tree leave nodes with SVM algorithms, the algorithm makes it possible for test-accuracy-rate-oriented joint optimization of decision tree and SVM parameters. This model is tested and validated with the actual data.3. This platform is designed and achieved based on Weka technology. The system includes integrated data warehouse, data mining and support vector machine (SVM) intelligent information processing technology, which is to build a system for the massive core business data, with ability to comprehensively analyze and process data mining. In this platform, data preprocessor, main algorithm model, meta-algorithm description Script Interpreter model and UI model are included and implemented.This system has now been accomplished, and experiments to predict the Probability of Default(PD) of customers show that the SDT-SVM is robust and indicates that the platform is reliable and trustworthy.