The financial fraud detection using machine learning

Abstract

The financial fraud detection system using machine learning is a modern tool for ensuring information security in financial institutions and commercial organizations. The relevance of this work is due to an increase in trends in the development of the use of cashless transactions, together with an increase in criminal offenses related to payment card fraud. An analysis of the research of scientists on this topic is provided and it shows that they cover the individual components of building a financial fraud detection system, but do not describe the complete cycle of the development and implementation of such a system. The two fundamentally different approaches to identifying financial fraud are considered – based on rules and based on machine learning tools. The advantage of using machine learning tools is substantiated in the context of improving the usability of the system, increasing the accuracy of fraud detection and possible integration with behavioral analytics systems. In this paper, the problem of detecting financial fraud with payment cards is formalized in machine learning terminology. The choice of the mathematical apparatus for the functioning of the model of detecting financial fraud with payment cards is substantiated. Mathematical algorithms are adapted to solve the problem of transaction classification and a step-by-step algorithm for the implementation of this machine learning task is given. The technical implementation of the system for detecting financial fraud with payment cards based on Microsoft Azure cloud services is developed and substantiated. The effectiveness of the proposed system for detecting fraudulent transactions is assessed, where sensitivity and specificity are selected as the criteria for efficiency being generally accepted indicators in machine learning theory.