新能源领域专利转让网络中技术供需主体间交易机会预测  

Prediction of Transaction Opportunities among Technology Supply and Demand Subjects in Patent Transfer Networks of New Energy Field

作　　者：武玉英[1] 张博闻 何喜军[1] 蒋国瑞[1] Wu Yuying;Zhang Bowen;He Xijun;Jiang Guorui(School of Economics and Management,Beijing Technology of University,Beijing 100124)

机构地区：[1]北京工业大学经济与管理学院,北京100124

出　　处：《情报杂志》2018年第5期79-84,96,共7页Journal of Intelligence

基　　金：国家自然科学基金项目"京津冀都市圈现代制造业生产要素协同创新研究"(编号:71371018);北京市社会科学基金项目"资源与环境约束下北京新能源汽车产业发展对策研究"(编号:15JGB124);北京市自然科学基金项目"要素异质性视角下京津冀现代制造产业转移路径研究"(编号:9172002)研究成果之一

摘　　要：[目的/意义]在专利转让网络中,通过挖掘供需主体信息及网络结构对交易机会的预测,有助于提升专利的转化效率。[方法/过程]以新能源领域为例,采集2012-2016年专利转让数据,构建专利技术交易加权有向网络。在转让数据统计分析与网络结构研究基础上,计算并筛选网络节点结构相似度指标,利用LDA主题模型结合余弦相似度方法计算节点内容相似性指标,通过熵权法将结构与内容指标融合,预测网络中专利技术供需主体间交易机会及专利技术流动方向。[结果/结论]结果表明:RA相比其他9个结构指标其预测精度更高,将RA与内容相似性指标融合后,预测精度略有上升,预测结果相比单一结构指标,更全面的挖掘出可能发生交易的供需主体对,结果更具多样化和合理性,在实践中更具有辅助决策的价值。[Purpose/Significance]In the patent transfer network,the prediction of transaction opportunities can help improve the efficiency of patent transformation by mining the information of supply and demand entities and the network structure.[Method/Process]Taking the newenergy field as an example,the patent technology transaction weighted and directed networks are constructed based on the patent transfer information collected from 2012 to 2016. On the basis of the statistical analysis of transfer data and the research on the structure of the networks,we calculate and select the structural similarity index of network nudes,combining the topic model of Latent Dirichlet Allocation（LDA） and the method of cosine similarity to confirm the nude content similarity index. We use the entropy method to carry out the structure and content similarity fusion to achieve the link prediction of inter-subject technical transaction opportunity and direction.[Result/Conclusion]The results showthat： RA can get the best prediction accuracy compared with other 9 structural similarity indexes; the prediction accuracy is slightly increased after the fusion of RA and the content index; compared to the single structure index,the supply and demand subjects making a pair likely to trade with each other identified are more comprehensive,the prediction results are more diversified and reasonable and are more valuable in decision support practice.

关 键 词：技术交易 加权有向网络 结构相似性 内容相似性 链路预测 

分 类 号：G306[文化科学]