An Application of Support Vector Machines in Small-Business Credit Scoring

This paper aims to apply a relatively new learning algorithm, support vector machines (SVM), to the credit scoring problem in a small-scale student dress wholesale company. Because most of the customers are minor small businesses that do not disclose financial information, it is almost impossible to obtain their financial data. So we propose an approach to assess the customers´ credit only based on daily transaction data such as sales, payments by customers, amount of overdue payment, etc. We provide the model of SVM for the credit scoring problem and discuss the appropriate choice of kennel functions and their parameters. We confirm the performance and effectiveness of the SVM model by applying it to the real problems of the company and comparing it with discriminant analysis.