문맥 기반 임베딩을 활용한 나노기업 자동 분류 연구 : 경상남도 사례를 중심으로

초록

This study aims to develop a model that automatically classifies nano-related firms within the manufacturing sector in Gyeongsangnam-do using SBERT, a context-aware sentence embedding model, and to analyze their spatial distribution characteristics. To address the limitations of subjective classification and the Korean Standard Industrial Classification (KSIC), a context-based embedding approach was applied to nano- technology definition texts and KSIC detailed classification descriptions, and cosine similarity was calculated to identify nano-related KSIC codes. Applying these codes to a dataset of 42,266 manufacturing firms in the region led to the identification of 2,181 nano-enterprises, which were found to be concentrated primarily in Changwon and Gimhae, with a particularly high presence in the nano-energy sector. This study presents a quantitative and automated classification framework that provides an objective foundation for policy-making and industrial analysis, and demonstrates the applicability of AI-based classification systems in public administration.

키워드