Hidden Markov Models (HMMs) has become increasingly popular in the last few decades due to its very rich mathematical structure and therefore forming the theoretical basis for use in a wide range of real-life applications such as in speech and image recognition, motion analysis in videos, bio-informatics among others. However, an effective optimization of the parameters of these Models for enhanced performance has remained computationally challenging and there is no generally agreed method that can guarantee best performance within reasonable computing time. Another significant challenge with the application of Machine learning algorithms to anomaly/fraud detection is the high number of false positives and negatives especially in the presence of highly class-imbalanced data sets. Designing an accurate efficient real-time Fraud Detection System (FDS) that is low on false positives and negatives but detects fraudulent activities effectively is essential. In this research, a hybrid algorithm comprising the Particle Swarm Optimization (PSO), BaumWelch (BW), and Genetic algorithms (GA) is proposed and implemented for optimizing the parameters of HMMs. A framework based on HMMs, modified Density Based Spatial Clustering of Applications with Noise (DBSCAN) and Synthetic Minority Oversampling Technique (SMOTE) is also implemented to effectively detect real-time electronic Banking fraud. An enhanced multi-layer HMM is proposed and implemented to further reduce the false positives and negative rates. Simulation results demonstrates that, the proposed hybrid optimization algorithm overcomes the weaknesses of the slow convergence of PSO whilst enabling the BW to achieve a global optimal solution. It also improves the performance of the GA by reducing its search space for optimal performance. Using highly imbalanced datasets, the proposed system performed relatively better when compared to some common approaches in literature in terms of precision, recall, F1-Scores and convergence rates. An improved multi-layer HMM proposed by the study also performed better with enhanced training and detection times as compared to other techniques widely used in literature.
A., A (2024). ENHANCED HIDDEN MARKOV MODELS (HMMs) FOR REAL-TIME FRAUD DETECTION IN ELECTRONIC BANKING. Afribary. Retrieved from https://track.afribary.com/works/enhanced-hidden-markov-models-hmms-for-real-time-fraud-detection-in-electronic-banking
A., Abubakari "ENHANCED HIDDEN MARKOV MODELS (HMMs) FOR REAL-TIME FRAUD DETECTION IN ELECTRONIC BANKING" Afribary. Afribary, 16 Jul. 2024, https://track.afribary.com/works/enhanced-hidden-markov-models-hmms-for-real-time-fraud-detection-in-electronic-banking. Accessed 24 Dec. 2024.
A., Abubakari . "ENHANCED HIDDEN MARKOV MODELS (HMMs) FOR REAL-TIME FRAUD DETECTION IN ELECTRONIC BANKING". Afribary, Afribary, 16 Jul. 2024. Web. 24 Dec. 2024. < https://track.afribary.com/works/enhanced-hidden-markov-models-hmms-for-real-time-fraud-detection-in-electronic-banking >.
A., Abubakari . "ENHANCED HIDDEN MARKOV MODELS (HMMs) FOR REAL-TIME FRAUD DETECTION IN ELECTRONIC BANKING" Afribary (2024). Accessed December 24, 2024. https://track.afribary.com/works/enhanced-hidden-markov-models-hmms-for-real-time-fraud-detection-in-electronic-banking