ABSTRACT: As the use of the internet is growing exponentially, more and more businesses such as the financial sector are initiating their services online. Accordingly, financial fraud is increasing in number and forms around the world, which results in financial losses which make financial fraud a major problem. Unauthorized access and immense regular attacks are examples of threats that should be detected by means of financial fraud detection systems. Machine learning and data mining techniques have been extensively used to tackle this problem over the past few years. However, these methods still need to be improved in terms of fast computation, dealing with huge data, and identifying the unknown attack patterns. Therefore, in this paper, deep learning-based method is implemented for the detection of financial fraudulence based on the Long Short-Term Memory (LSTM) technique and Bidirectional encoder representation from transformers (BERT) . This model is aimed at enhancing the present detection techniques as well as enhancing the detection accuracy in the light of big data. To evaluate the proposed model, a real dataset of credit card frauds is utilized and the results are compared with an existing deep learning model named spiking neural network and some other machine learning techniques. The experimental results illustrated a perfect performance of LSTM where it achieved 99.95% of accuracy.
KEYWORDS : Fraud ; fraud detection; deep learning; long short-term memory
|
DOI:
10.17148/IJARCCE.2022.11610
