Answers are grounded in this paper and cite their sources.
Comparative analysis of machine learning and transfer learning models for classifying business descriptions, suggesting enhanced efficiency in economic statistics.
The Indonesian Standard Classification of Business Fields (KBLI) is essential for economic statistics, yet manual classification of business descriptions to five-digit KBLI codes is time-consuming and prone to inconsistencies. This study aims to develop and compare machine learning (Support Vector Machine and Random Forest) and transfer learning (IndoBERT) models for automating KBLI classification, supported by the preparation of synthetic and real-world datasets for model training. The synthetic data were generated using large language models, validated through human majority voting and complemented with realworld data from the National Labor Force Survey (Sakernas) and the Micro and Small Industry Survey (IMK). The findings indicate that Fine-tuned IndoBERT achieved superior performance, achieving an F1-score of 92.99% and an accuracy of 93.40% on synthetic data, alongside top-1, top-5, and top-10 accuracies of 32.93%, 54.71%, and 63.24% on real-world data. The deployment of fine-tuned IndoBERT as a RESTful API demonstrates its scalability and efficiency, presenting a reliable solution for large-scale KBLI classification in official statistics.
No takes yet. Share an insight, caveat, or question.
Amnur et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: