Abstract

Deep Learning Model for ITRS Data Quality Enhancement

The contemporary landscape of global finance and foreign exchange liberalization, on the one hand, could facilitate business operations. On the other hand, they bring along with them challenges for data compilation, economic analysis, and policy formulation. Regarding data compilation, one critical challenge involves handling unstructured transaction purpose descriptions as reported by commercial banks through the International Transaction Reporting System (ITRS). This research aims to address this challenge by transforming unstructured transaction description into structured format, enabling more accurate reclassification of cross-border transactions in the balance of payments (BOP).