基于综合引用网络的专利价值评价研究

Research on Patent Value Evaluation Based on Comprehensive Citation Network

目前，利用专利引文进行专利价值评价的方法，通常是采用直接引文的被引频次作为评价指标，该方法会导致对于整体网络引用关系的不完整理解。本文基于矩阵转化方法对四种单一专利引用关系（直接引用、间接引用、耦合、共引）进行合并、筛选、重组的基础上，提出了一种适合专利价值评价的专利综合引用网络构建方法。通过对光盘技术专利数据的实证研究发现：该网络在保留直接引用网络价值评价优势的基础上，在拓扑属性、度分布特征以及引用时滞等网络特性上比直接引用网络具有更好的性能；利用专利家族和专利持续时间数据进行验证后发现：专利综合引用网络在有价值专利的筛选方面，比直接引用网络更加全面且准确度更高，因此，是一种适合专利评价的专利综合引用网络分析方法。

Currently, most approaches to patent value estimation are based on only direct patent citation, which leads to an incomplete understanding of the nature of linkages. In this paper, four kinds of single patent citation relations： direct citation, indirect citation, coupling and co-citation networks are combined, selected and recomposed by matrix translation method, then a method for building comprehensive patent citation （CPC） network to evaluate patent value is established. The empirical research on optical disc technology related patents have been carried out, and results show the CPC network keeps the advantages of the direct citation, and has better performance on topological structure, centrality distribution and citation time lag than that of direct citation network. The results also show the CPC network is more accurate on patent value evaluation and can be used to select more valuable patents.