Detection of credit card fraud using optimization algorithms

Credit card fraud is a very common problem today, with the spread of e-commerce applications. Especially during the COVID-19 period, people do their shopping and all their transactions via the Internet while living under quarantine at home. Most people carry out their transactions quickly and practically through shopping sites using credit cards or mobile banking applications. However, in addition to these advantages, transactions made as a result of stealing credit card information, copying the credit card and using credit cards belonging to others pose some legal and economic problems. The aim of this study is to develop genetic algorithm methods based on roulette and tournament selection instead of random selection in identifying and classifying these illegal credit card fraud transactions. A data set consisting of 28 attributes and 284807 credit card transactions was used in the study. Fraud in credit card transactions was predicted using genetic algorithm and artificial bee colony and the results were compared according to roulette and tournament selection. Optimization methods such as genetic algorithm and artificial bee colony developed based on roulette and tournament selection were analyzed separately according to linear, quadratic and exponential functions. Using the linear quadratic and exponential functions for the test data, the genetic algorithm based on roulette selection showed success rates of 98.6%, 98.46% and 98.6% in identifying credit card fraudulent transactions, respectively; With the genetic algorithm based on tournament selection, it shows 98.53%, 98.33% and 98.66% success. In addition, success results were calculated with different evaluation criteria such as precision, recall and F1-Score, and complexity matrices were presented for each proposed method.