基于机器学习的高质量专利特征组合挖掘  

Mining Features Combination of High-quality Patents by Machine Learning

作　　者：周磊 王婧怡 黄彩云 余玲珑 ZHOU Lei;WANG Jing-yi;HUANG Cai-yun;YU Ling-long(Accounting College,Wuhan Textile University,Wuhan Hubei 430200,China;School of Management,Wuhan Textile University,Wuhan Hubei 430200,China)

机构地区：[1]武汉纺织大学会计学院,湖北武汉430200 [2]武汉纺织大学管理学院,湖北武汉430200

出　　处：《武汉纺织大学学报》2021年第3期67-71,共5页Journal of Wuhan Textile University

基　　金：国家社会科学基金青年项目(19CTQ030).

摘　　要：挖掘高质量专利的特征组合有助于揭示专利价值形成机制,进而引导专利质量整体提升。以中国专利奖获奖发明专利为样本集,从技术质量、法律质量和经济质量等指标建立高质量专利评价指标体系,进而利用决策树模型抽取出9条区分金奖发明和优秀发明的知识规则。发现法律质量是高质量专利的第一要素,高权利项可作为识别金奖发明的唯一特征;较高的权利项与专利族、被引频次的组合可以识别金奖发明;权利项偏低时,专利需满足多个技术质量指标阈值才能认定为金奖发明。Mining features combination of high-quality patents is helpful to revealing the mechanism forming patent value and improving the overall quality of patents.Treating award-winning inventions of China Patent Award as the sample,a high-quality patent evaluation index system is established including technical quality,legal quality,and economic quality.And then,9 knowledge rules are extracted to differ gold award inventions from excellent award inventions by a decision tree model.Several findings are drawn from such 9 rules.Firstly,legal quality is the fundamental element for high-quality patent with the evidence that patent claim is regard as the individual feature of gold award invention.Secondly,the combination of larger patent claims,patent family and yearly citations refers to gold award invention.Thirdly,a gold award invention with smaller patent claims must reach the thresholds of multiple technical quality index simultaneously.

关 键 词：专利质量 机器学习 决策树 专利权利项 中国专利奖 

分 类 号：F204[经济管理—国民经济]