Analysis of the use of the supervised machine and deep learning techniques in the detection of financial fraud

Abstract

In the modern world, it is necessary to use techniques, methodologies, and actions in search of the integration of the various advances, tools, and current elements for joint work in solving the problems that affect the finances of organizations, since they make a business dynamic exist, creating economic value. Taking into account the above, this study analyzes the prevention of business fraud, through the use of automatic and deep learning techniques to generate prevention, treatment, and resolution of fraud carried out in financial systems. At the methodological level, information was obtained in databases at the documentary level, with reliable sources and case studies where the effectiveness of the use of the aforementioned techniques in the early detection of business fraud is tested.

The results obtained in the documents consulted express that the algorithms that are most effective in preventing these frauds are decision tree, C5.0-SVM, Naïve Bayes, and Random Forest, with percentages of 92%, and 83.15%, 80, 4%, and 76.7% respectively. In contrast to deep learning, the literature showed that by making use of neural arithmetic logic units and performing the correct classification of the iNALU and ReLU neurons, the percentage of effectiveness increases greatly.

In the final part of this document, results and conclusions are presented and consolidated, all within the framework of the topic addressed, in addition, the information compiled in this document is duly supported by the copyright to whom it corresponds.