识别高被引专利——基于稀有事件Logit与倾向得分匹配模型

Identifying Highly Cited Patents:Based on Rare Event Logit and Propensity Score Matching Model

通过对专利价值评估方法的研究,可以帮助企业识别有价值的专利,以高效保护企业的技术创新。文章运用社会网络分析方法,从Thomson Innovation数据库选取LED行业专利作为样本,根据专利被引证数计算得出专利H指数,以此作为专利价值的代理变量,并以专利的社会网络指标作为自变量。传统logit回归算法在参数估计、统计推断、选择性偏差和内生性问题方面都有可能存在缺陷。文章在logit回归模型基础上,运用稀有事件(Rare Events)logit回归模型和倾向得分匹配(Propensity Score Matching)模型对专利价值的影响因素进行系统分析。研究发现,在3种模型下,外向程度中心性、专利PageRank值、中间中心性和特征向量中心性与专利价值均呈正向关系。

As intangible assets,patents can improve the competitiveness of enterprises and bring huge potential value to them.Therefore,research on the methods of patent value assessment can help enterprises to identify valuable patents,effectively protect their technological innovation,and obtain higher quality and more valuable patents.Using social network analysis method,LED industry patents up to 2016 have been selected from Thomson Innovation database as samples,and the H-index of these patents is calculated based on the times cited.Then,the index is used as the proxy variable of patent value while the social network index of patents is used as the independent variable.Traditional logit regression may have inherent weakness in terms of parameter estimation,statistical inference and selection bias.So,the rare events logit regression model and the tendency to propensity score matching model are used to analyze the factors influencing the value of patent system.It is found that extraversion centrality,patent PageRank value,intermediate centrality and eigenvector centrality are positively correlated with patent value in three models.The results of this study are of guiding and referential significance for enterprises to identify those factors affecting patent citation evaluation and for them to make technological inventions.