[Purpose/Significance]This paper builds a disruptive indicator system based on patent terminologies to identify and predict disruptive technologies,and selects relevant patents in the field of solar photovoltaics for empirical study.[Method/Process]This paper built a disruptive indicator system with 5 indicators from patent full texts by consid-ering two dimensions of technological innovation and technological influence based on patent terminologies.The process transformed the identification problem into a classic binary classification problem through the natural language processing techniques to train a machine learning model,and introduced decision curve analysis to determine the optimal classifi-cation threshold for the problem of unbalanced samples.[Result/Conclusion]The empirical study in the field of solar photovoltaic identifies 91 disruptive patents among 2 196 patents,which gets a better predictive performance and verifies the validity of the indicators.In summary,the proposed approach provides a new perspective for future research on how to predict disruptive technologies,measure patent value,and identify promising technologies.
关键词
颠覆性技术/指标体系/技术创新性/技术影响力/专利全文本
Key words
disruptive technology/indicator system/technological innovation/technological influence/patent full text
维普
万方数据