Integrating knowledge graphs and LLM for Indonesia HS code classification: A focus on heavy equipment parts

Accurate Harmonized System (HS) code classification is crucial for regulatory compliance and cost efficiency in global given the dynamic tariff policies in Indonesia and globally. This study introduces a framework integrating Knowledge Graphs (KGs), Large Language Models (LLMs), and Machine Learning (ML) to improve HS code classification for heavy equipment parts. ML models predict classifications using historical data, while KGs encode domain-specific regulations for consistency LLMs provide transparent, human-readable explanation and auditability in the classification process. The integrated framework achieved high accuracy (95%) and demonstrated excellent explainability through Retrieval-Augmented Generation Assessment (RAGAS) with near-perfect performance in context recall and precision.