AN EMPIRICAL STUDY ON AN AI-BASED MODEL FOR ENHANCING CUSTOMS TAX SUPERVISION AND COMBATING TAX EVASION

Abstract

This research creates and assesses a data-driven model for improving the supervision of customs taxes and the fight against tax evasion. The model combines customs declaration data and analytics for risk assessment, as well as identifying anomalies and prioritizing inspections for customs administrations. The model, built from analytics, and a traditional rule-based monitoring system are compared across several performance dimensions using historical customs declaration data. The findings show that the new model attains better detection precision, greatly improves the accuracy of classifications to reduce false positives, and improves the targeting of detected irregularities to the high-risk declaration categories. Furthermore, the results indicate that the model improves the effectiveness of inspections and the administrative burden is lessened, all without the loss of revenue protection. The enhancements are not simply measures of technical improvements, but are operational and managerial enhancements that are relevant for contemporary customs administrations.