专利视域下融合协同过滤与链路预测的企业潜在合作关系预测研究被引量：2

Research on Predicting Potential Firm Collaborations by Integrating Collaborative Filtering and Link Prediction in the Patent Perspective

下载PDF

导出

摘要 [研究目的]融合协同过滤与链路预测对“企业—专利”关系进行系统而细致的逻辑表达,继而精准有效地识别企业潜在合作关系,有助于企业聚合技术资源并推动产业创新化进程。[研究方法]基于非晶合金专利数据构建合作网络,通过Pearson相关系数计算专利文本的内容相似性,利用SM算法计算IPC号的类别相似度,使用SimRank指标和RA指标计算链路预测的路径相似度,随后借助权重预测算法构建以三者为基础的融合加权指标。将融合指标作为协同过滤推荐算法的输入,进而预测企业潜在合作伙伴。[研究结论]研究表明,融合协同过滤和链路预测的“链路推荐”算法预测平均准确率达到93.21%,较之传统的协同过滤算法提升了2.42%左右。在非晶合金领域企业潜在合作关系预测过程中,由“链路推荐”算法得出的推荐结果能够为企业合作提供决策与参考。 [Research purpose]This study aims to provide a systematic and detailed logical expression of the"firm-patent"relationship by integrating collaborative filtering and link prediction.It further aims to accurately and effectively identify potential cooperative relationships among firms,which can help firms aggregate technological resources and promote industrial innovation processes.[Research method]A cooperative network is constructed based on non-crystalline alloy patent data.The content similarity of patent texts is calculated using pearson correlation coefficient,the category similarity of IPC codes is calculated using the SM algorithm,and the path similarity of link prediction is calculated using SimRank and RA metrics.A fusion weighted index is then constructed based on these three metrics using a weight prediction algorithm.The fusion index is used as input for the collaborative filtering recommendation algorithm to predict potential firm partnerships.[Research conclusion]The study shows that the"link recommendation"algorithm,which combines collaborative filtering and link prediction,achieves an average accuracy rate of 93.21%,an improvement of approximately 2.42%compared to traditional collaborative filtering algorithms.In the process of predicting potential cooperative relationships among firms in the field of non-crystalline alloys,the recommended results obtained from the“link recommendation”algorithm can provide decision-making and reference for firm collaborations.

作者周志刚窦路遥李毅 Zhou Zhigang;Dou Luyao;Li Yi(Institute of Information Science,Shanxi University of Finance&Economics,Taiyuan 030006)

机构地区山西财经大学信息学院

出处《情报杂志》 CSSCI 北大核心 2023年第12期144-152,共9页 Journal of Intelligence

基金中国高等教育学会专项课题“大数据技术与应用产教融合实践与创新”(编号:21CJZD06) 山西省哲学社会科学规划课题“山西省中小企业间数据融合安全共享机制研究”(编号:2022YY097)的研究成果。

关键词协同过滤链路预测专利文本专利合作网络潜在合作关系预测 collaborative filtering link prediction patent text patent cooperation network potential cooperation relationship prediction

分类号 G251 [文化科学—图书馆学]