[Research purpose]The development of science,technology,and society has driven the integration of theories,methods,and techniques across diverse fields.Cross-disciplinary collaboration is increasingly becoming the predominant form of cooperative innovation.Identifying and accurately pinpointing potential partners for such collaborations has emerged as a pressing challenge.[Research method]This paper introduces a novel method for the identification of cross-domain collaboration partners of inventors utilizing Graph Convolution-al Networks(GCNs).By adopting a multi-dimensional feature perspective,we harnessed features from cooperation relationships,abstract textual content,and domain-specific data inherent within inventor patent records.Through the application of GCNs,we were able to sys-tematically identify and predict potential collaborators for inventors,strategically developing both intra-domain and inter-domain indices for precise identification of cross-domain collaborative partnerships.[Research conclusion]Through comparative experiments,it has been demonstrated that leveraging the GCNs in conjunction with three-dimensional features—collaboration relationship,field information,and abstract textual features—can significantly enhance model accuracy.Identifying cross-disciplinary collaborators fosters cross-field col-laborations and knowledge transfer,leading to more innovative and forward-looking outcomes.
维普
万方数据