article · Machine Learning with Applications
Online credit card fraud remains a severe challenge for e-commerce platforms and financial institutions, as misclassification leads directly to financial loss. To address this issue, an emerging online learning method known as Hierarchical Temporal Memory based on Cortical Learning Algorithms (HTM-CLA) was evaluated alongside other machine learning techniques. The investigation compared HTM-CLA against an Artificial Neural Network trained using Simulated Annealing (SA-ANN) and a Long Short-Term Memory network (LSTM-ANN). Evaluations were carried out using simulation tests on two established benchmark datasets, namely the Australian and German credit card fraud data, measuring success through an average classification performance ratio metric. The simulation findings revealed that HTM-CLA demonstrated competitive performance when compared to the SA-ANN model. Furthermore, HTM-CLA substantially exceeded the performance of the deep recurrent LSTM-ANN system across the tested benchmark datasets, achieving an advantage by a factor of two to one.
Online fraud inflicts major financial damage on digital merchants and banking institutions. Improving automated anomaly detection helps reduce misclassification errors during payment processing. Demonstrating that alternative learning architectures such as Hierarchical Temporal Memory can rival or outperform conventional recurrent neural networks highlights practical routes for improving fraud detection efficiency.
The evaluated models target fraud detection systems for financial institutions and e-commerce merchants seeking to minimise transactional losses. Because the findings are based on simulations across standard Australian and German benchmark datasets, the technology is at an applied research stage. Further operational validation on live, streaming transaction data would be required before integrating these algorithms into commercial fraud prevention software.
AI-generated from the published abstract. Always read the original work before citing.
The problem of misclassification has always been a major concern in detecting online credit card fraud in e-commerce systems. This concern greatly poses a significant challenge to financial institutions and online merchants with regards to financial loss. This paper specifically compares an Artificial Neural Network trained by the Simulated Annealing technique (SA-ANN) with a proposed emerging online learning technology in anomaly detection known as the Hierarchical Temporal Memory based on the Cortical Learning Algorithms (HTM-CLA). Comparisons are also made with a deep recurrent neural technique based on the Long Short-Term Memory ANN (LSTM-ANN). The performances of these systems are investigated on the basis of correctly classifying credit card fraud (CCF) using an average classification performance ratio metric. The results of simulations on two CCF benchmark datasets (the Australian and German CCF data) showed promising competitive performance of the proposed HTM-CLA with the SA-ANN. The HTM-CLA also clearly outperformed the LSTM-ANN in the considered benchmark datasets by a factor of 2:1.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.1016/j.mlwa.2021.100080
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
