FRAUD DETECTION WITH NATURAL LANGUAGE PROCESSING

Abstract

Automated fraud detection may aid companies in protecting user accounts, a job that is particularly difficult owing to the scarcity of proven fraudulent transactions. A significant portion of the existing literature primarily addresses credit card theft while neglecting the emerging field of internet banking. Nevertheless, there is a dearth of readily accessible data for both. The absence of readily accessible data impedes the advancement of the field and restricts the exploration of possible remedies. This work accomplishes three main objectives. Firstly, we present FraudNLP, which is the initial anatomized dataset accessible to the public for online fraud detection. Secondly, we evaluate various machine and deep learning techniques using multiple assessment metrics. Lastly, we demonstrate that online actions adhere to patterns similar to natural language, making them amenable to successful analysis using natural language processing methods.