Fuzzy logic and Takagi-Sugeno Neural-Fuzzy to Deutsche bank fraud transactions

摘要

This article proposes suitable solution to detect fraud via fuzzy logic followed by Neural-fuzzy Takagi-Sugeno training method. In order for the fraud to be detected through fuzzy logic, there should be some rules stemmed from experience of the experts. These rules are expressed through information that could be registered for a given card. To come up with the fuzzy deduction, membership functions needed to be expressed over the specified input range. This issue is one of the problems of fuzzy logic. To solve this problem, fuzzy logics were established and Mamdani deduction engines were utilized as a result of which suitable responses were presented for fraud detection via Neural-fuzzy method. Despite the fact that the problem inputs were highly linear, Neural-fuzzy training was able to cope with the problem and present a suitable trained system. In other words, Neural-fuzzy training method is employed in order to optimize the fuzzy logic membership functions based on the data. Outcomes of the Neural-fuzzy training were quite satisfactory and highly precise. Thus, utilizing the research findings, Neural-fuzzy training method is proposed for upgrading fraud detection in the banking system of our country.