Payment card fraud detection using machine learning

Payment card frauds cost billions of dollars to all card issuers every year. Therefore, banks make more effort every year to improve the methods of analyzing data to effectively detect card frauds. There are various techniques for detecting fraud transactions but the most effective and used ones are deep learning methods. Deep learning as one of the subsets of machine learning is becoming the most popular research point. Deep learning uses artificial neural networks. Artificial neural networks are flexible and self-adaptive to solve complex problems that are difficult to describe with a mathematical model. The purpose of this work is to implement a neural network algorithm able to analyze a credit card fraud data set originated from one local Czech bank. The type of neural network used is multilayer perceptron. Neural networks with different parameters of layers, neurons, activation functions and optimizers were developed. To compare the models metrics Recall, Precision, F1-score and AUROC were used.