A Fraud-Detection Fuzzy Logic Based System for the Sudanese Financial Sector

Abstract

Financial fraud considered as a global issue that faces the financial sector and economy; as a result, many financial institutions loose hundreds of millions of dollars annually due to fraud. In Sudan, there are difficulties of getting real data from banks and the unavailability of systems which explain the reasons of suspicious transaction. Hence, there is a need for transparent techniques which can automatically detect fraud with high accuracy and identify its causes and common patterns. Some of the Artificial Intelligence (AI) techniques provide good predictive models, nevertheless they are considered as black-box models which are not easy to understand and analyze. In this paper, we developed a novel intelligent type-2 Fuzzy Logic Systems (FLSs) which can detect fraud in debit cards using real world dataset extracted from financial institutions in Sudan. FLSs provide white-box transparent models which employ linguistic labels and IF-Then rules which could be easily analyzed, interpreted and augmented by the fraud experts. The proposed type-2 FLS system learnt its fuzzy sets parameters from data using Fuzzy C-means (FCM) clustering as well as learning the FLS rules from data. The proposed system has the potential to result in highly accurate automatic fraud-detection for the Sudanese financial institutions and banking sectors.