특허정보와 기계학습을 활용한 산업기술수준 측정 방법 연구

Accurate measurement of technology level is required as basic information for establishing technology policy or strategy. However, technology level survey from experts may lack objectivity and consume considerable time and costs. Therefore, this study was conducted to derive a methodology to measure the level of technology objectively and easily by using patent information. We attempted to measure the industrial technology level from patent information by linking survey results with the patent competitiveness data of 5 major countries in 80 industrial technology fields. Using various patent indices as independent variables and the technology level survey result as a dependent variable, linear regression analysis was performed, identifying the influence of each index on the determination of technology level. Next, the technology level was predicted using artificial neural networks, random forest, and XGboost, confirming the better performance of machine learning than linear regression method. This study is meaningful in that it developed a technology level evaluation methodology with improved accuracy by using various patent indices and machine learning. And we expect our research would be used as a tool to supplement the expert survey method.