ADVANCED JOURNAL OF ENGINEERING AND SCIENTIFIC APPLICATIONS

https://ajesa.com.ng/index.php/AJESA/issue/feed ADVANCED JOURNAL OF ENGINEERING AND SCIENTIFIC APPLICATIONS 2026-01-13T01:10:40+00:00 Engr. Dr. Fidelis C. Obodoeze PhD, MNSE, NIEEE, FIIA drfcobodoeze@gmail.com Open Journal Systems <p>ADVANCED JOURNAL OF ENGINEERING AND SCIENTIFIC APPLICATIONS (AJESA) is an academic/technical Journal of engineering and applied sciences published by the Nigerian Institute of Electrical and Electronic Engineers (NIEEE) Awka Chapter covering all fields of Engineering and Engineering Sciences such as Mechanical Engineering, Safety and Fire Engineering, Civil Engineering, Electrical and Electronic Engineering, Computer Engineering, Communication Engineering, Computer Science, Software Engineering, Cybersecurity, Information Technology, Entrepreneurship Studies, Chemical Engineering, Petrochemical Engineering, Water Engineering, Highway Engineering, Materials and Metallurgical engineering, current trends in Artificial Intelligence, Robotics and Machine Learning, Polymer Engineering, etc.</p> <p>The Journal is an open-access online journal publishing double-blind peer-reviewed research papers in every area of engineering and scientific applications related to engineering and engineering sciences and sub-fields. It is published every month within two-months interval, but paper submission is open every month.</p> https://ajesa.com.ng/index.php/AJESA/article/view/36 MACHINE LEARNING APPROACHES TO FRAUD DETECTION AND RISK ANALYTICS IN GSM-BASED FINTECH SYSTEMS IN NIGERIA 2026-01-13T01:10:40+00:00 Joseph Osahon Idemudia jidemudia@akanuibiampoly.edu.ng Taiwo Adisa Oyeniran taoyeniran@akanuibiampoly.edu.ng

The increasing adoption of mobile phone numbers as primary identifiers in financial transactions has significantly transformed the operations of financial technology (FinTech) companies in Nigeria. Leveraging Global System for Mobile Communications (GSM) numbers as account identifiers has enhanced access to digital financial services while introducing challenges related to security, data management, and analytics. This study investigates the application of data mining and machine learning techniques in GSM-based FinTech systems, with emphasis on fraud detection, anomaly detection, customer segmentation, and credit risk assessment. A synthetic yet statistically representative GSM-linked transaction dataset was developed to reflect real-world FinTech operations under regulatory and ethical constraints. Controlled modeling experiments were conducted using supervised and unsupervised learning techniques. Experimental results show that the Random Forest classifier achieved an accuracy of 96.8%, precision of 94.5%, recall of 92.7%, and an F1-score of 0.935, substantially outperforming traditional rule-based approaches. Receiver Operating Characteristic analysis yielded a high area under the curve exceeding 0.94, indicating strong discriminative performance. Anomaly detection models effectively identified SIM-swap and account takeover fraud, achieving detection rates above 90% with average detection delays below two minutes. Unsupervised clustering revealed four distinct customer segments, including a high-risk group representing approximately 12% of users. In addition, predictive credit risk models achieved 92.1% accuracy and reduced simulated loan approval time by 37%. Feature importance analysis identified GSM-specific behavioral indicators—such as transaction value, SIM change events, and location instability—as key predictors of fraudulent and high-risk behavior. Overall, the results demonstrate that data mining significantly enhances security, operational efficiency, and financial inclusion in GSM-based FinTech systems.