A Novel Trio-Hybrid for Detecting Fraudulent Credit Card Transactions

In this era of digitization, credit card frauds shake down the spirits of not only the customers but also the merchants, which incurs a loss of billions of dollars globally. To combat such frauds, a robust and responsive system is needed that can flag the fraudulent transaction instantly before it happens. The existing systems are great at detecting and battling with fraud after it has happened but slouch in case of prevention of such crimes. They aren’t good¬¬ at optimization and also struggle in terms of response time. The inefficiency of existing systems is attributed to either working on a single machine learning technique, or just combining two of them. We present a Trio-Hybrid of K-means, Genetic algorithm, and artificial neural network approaches to deal with the aforementioned problems. The K-means algorithm helps in reducing the training time of neural networks and the genetic algorithm helps in feature selection to prevent the neural network from being over-trained, thereby reducing the cost of the system. We leverage the benefits provided by these three techniques and put them together into a trio for the first time and achieve an accuracy of 99.94% in detecting the fraudulent credit card transactions.