MACHINE LEARNING MODELS FOR RISK MANAGEMENT IN NIGERIAN CUSTOMS: AN INVESTIGATIVE PERFORMANCE ANALYSIS

Abstract

Customs administrations utilize risk analysis to identify which people, products, and modes of transportation should be scrutinized and to what extent. Risk analysis and risk assessment are analytical techniques for determining which risks are the most significant and should be treated first or have corrective action performed first. Several ML models were investigated to determine the suitable model for custom data. This is necessary due to the unavailability of such research work. The Machine Learning (ML) models considered are; Support Vector Machine (SVM), Decision Tree (DT) classifier, K-Nearest Neighbor (KNN), Ensemble and Discriminant analysis classifiers. The dataset was collected and pre-processed. The Models were trained and tested using 70% of data for training and 30% for testing. The result shows that the ensemble models produce the highest accuracy of 66.6% for Boosted Trees classifier when compared with the other models. The medium and coarse tree produced an accuracy of 66.1%. This shows that the tree-based algorithms performs averagely better than others and recommended for further exploration.