Fuzzy Support Vector Machine Based Outlier Detection for Financial Credit Score Prediction System

Abstract

Economic Credit Scoring (CS), which aids in calculating the credit worth of both individuals and companies, is regarded as one of the greatest study issues in the finance field. In the banking industry, data mining techniques are believed to be helpful since they help designers and developers create appropriate goods or services for customers with the fewest possible risks. Losses and loan cancellations, which are the major sources of hazards in the banking industry, are related to credit risks. A Support Vector Machine based architecture is presented in the current study for the financial credit score prediction system. However, the existing work tends to have increased computational overhead and that requires complete data for attaining required accurate rate. The system known as the Fuzzy Support Vector Machine based Outlier Detection System is introduced in the suggested research study to address this (FSVM-ODS). This study's first grouping of data items utilising a hybrid genetic algorithm with K-Means clustering algorithm is named (HKGA). The dataset must be gradually lowered in size, and calculation time must likewise be decreased. The Enhanced Z-score (EZS) outlier identification (OD) technique was employed in the second step to identify outliers in the dataset. Then, we use a customized beaver searching method to choose the database. For categorization of the datasets, a fuzzy support vector machine is utilised. The whole study project is carried out in the Matlab simulation environment, and it has been shown that the suggested technique achieves a higher outlier identification rate than the current methodology.