Credit Risk Prediction Model Using Support Vector Machine with Parameter Optimization in Banks

Abstract This research aims to determine the Support Vector Machine (SVM) model with Parameter Optimization in predicting loan worthiness to avoid the risk of bad credit at the Bank. Every bank tries to market financial loan products with very strict requirements. One of the requirements is that the company's financial reports must be healthy if it borrows money from a bank to develop the company's business. In the credit analysis process, there are 19 financial factors that must be measured from dozens or even hundreds of companies proposing financial loans, making it difficult for credit analysts to make decisions about whether these companies are worthy of borrowing or not. Therefore, this research was carried out by comparing the two models, namely SVM with parameter optimization and SVM with parameter optimization and Particle Swarm Optimization (PSO) to select the best model. The research results show that the Area Under Curve (AUC) criteria with validation number of folds (nof) = 10 and nof = 5 are 98.80% and 98.80%, meaning good and stable in the SVM model with parameter optimization. Meanwhile, the SVM model with parameter optimization and PSO has better AUC for validation nof=5 (99%) but for AUC with validation nof=10 (98.30%) it is less good.