The role of artificial intelligence in fraud analysis and prevention in Gabon

Yvan, Obiang Reliwa Placide and Xianwen, Fang and Mboungou, Marcel Merimee Bakala (2025) The role of artificial intelligence in fraud analysis and prevention in Gabon. International Journal of Science and Research Archive, 15 (3). pp. 803-812. ISSN 2582-8185

The study's overall objective is to showcase how an artificial intelligence (AI) capability can assist in making inroads on financial fraud mitigation and detection of fraudulent activity in Gabonese banking. The study conformed a series of different AI models: Decision Tree, Random Forest, Support Vector Machines (SVM) and, Autoencoders for anomaly detection with actual anonymized, transactional data from Gabonese banks. We managed to handle the extreme imbalance between the fraudulent (façade) transactions and legitimate (real) transactions and were able to standardize our data set with the Synthetic Minority Oversampling Technique (SMOTE) before testing the Model's decision-making capabilities. The outcomes of our discussions about the model's testing and performance suggested that, overall, the Autoencoder produced the strongest performance, achieving an F1-Score = 0.86, along with exhibiting strong TPS performance relative to an acceptable level of false negatives it produced. Random Forest, (F1-Score = 0.85) was not far below in performance suggesting the effectiveness of ensemble learning in this instance to map more complex patterns of fraud. Decision Tree and SVM also produced respectable scores with F1-Scores of 0.81 and 0.77 respectively. These results show, that AI has the potential to be a game-changer in fraud detection within Gabonese banking. The use of AI now provides decision-makers a new landscape to improve operations security, to decreased risk and potential financial losses, to positively impact customer confidence by detecting fraud while using data-based tools with rich machine learning capabilities in real-time. In addition, this study highlighted the importance of continuously testing and enhancing the model after the model has been delivered, with an ethical frame of reference that would protect against fair or practical outcomes of the model and the emergence of new types of fraudulent activity.