Predicting Customs Fraud Using Machine Learning and Mirror Analysis in Togo

Customs fraud is an inherent phenomenon of customs administrations and is most often responsible for undermining customs revenue collection. In an attempt to combat this phenomenon, customs administrations, particularly in developing countries, often conduct extensive and unstructured audits. This is not conducive to the fluidity of international trade. The objective of this study is to analyse the extent to which the use of machine learning and mirror analysis improves the identification of customs fraud, while preserving the objective of revenue mobilisation. Using data from the Togolese Revenue Authority and COMTRADE, the findings indicate that mirror analysis and machine learning can better enhance customs fraud detection. To this end, the study recommends the use of these tools in fraud detection.