摘要
发明人合作是专利合作的显著表现形式之一,通过构建合作网络,挖掘潜在的合作关系,有助于预测专利合作网络中发明人的未来合作趋势.考虑合作网络节点的位置信息和属性信息,首先,引入社交网络中的链路预测方法计算节点位置的相似性;其次,将发明人的研究方向作为节点的属性信息,分别运用Doc2vec和TF-IDF建立研究方向的向量耦合矩阵,计算发明人研究方向之间的余弦相似度,衡量发明人研究方向的相近性;最后,构建基于链路预测算法与发明人研究方向的混合算法,从而预测发明人的潜在合作关系.在国内知识图谱领域进行实证研究发现,TF-IDF在该领域的预测效果较好,并且在链路预测算法的基础上,通过融入研究方向相近性矩阵,预测精度得到了较好的提升.
Inventor cooperation is one of the obvious manifestations of patent cooperation.By building a cooperation network and digging out potential cooperation relationships,it is helpful to predict the future cooperation trend of in⁃ventors in the patent cooperation network.Considering the location information and attribute information of coopera⁃tive network nodes,first,the link prediction method in social network is introduced to calculate the similarity of node locations;then,the inventor's research direction is used as the attribute information of nodes,and Doc2vec and TF⁃IDF are used respectively.Establish the vector coupling matrix of research directions,calculate the cosine similarity between the inventors'research directions,and measure the similarity of the inventor's research directions;finally,construct a hybrid algorithm based on the link prediction algorithm and the inventor's research direction,so as to predict the inventors potential partnership.Empirical research in the field of domestic knowledge graph found that TF⁃IDF has a better prediction effect in this field,and based on the link prediction algorithm,the prediction accura⁃cy has been improved by incorporating the research direction similarity matrix.
作者
于凯
郭煜婕
YU Kai;GUO Yu-jie(School of Public Administration,Xinjiang University of Finance and Economics,Urumqi 830012,China;School of Information Management,Xinjiang University of Finance and Economics,Urumqi 830012,China)
出处
《云南民族大学学报（自然科学版）》
CAS
2024年第3期377-385,共9页
Journal of Yunnan Minzu University:Natural Sciences Edition
基金
新疆维吾尔自治区自然科学基金项目(2019D01A22)
新疆维吾尔自治区社科基金项目(21BTQ162).
