[Research purpose]Technology resource competition among countries has shifted from single technology competition to tech-nology cluster competition.Identifying technology cluster from the perspective of dynamic patent technology network can help enterprises and researchers to understand technology structure,explore technology opportunities and predict the future of technology,so that they can gain advantages in technology research and development and competition.[Research method]The research uses self-attention mecha-nism to learn the specific word vector of patent document,and integrates the prior information such as attribute technology factor and func-tional technology factor to generate the patent technology vector.After calculating the patent technology similarity,a dynamic patent tech-nology network is built,and the Louvain algorithm with node increment is further used to identify the technology cluster.[Research con-clusion]Based on this method,12 technology clusters such as data system,data equipment and data security,and 1 technology cluster fu-sion event are identified in an empirical study,which takes the e-commerce data technology in the United States as an example and further verifies the effectiveness of the method.
维普
万方数据