Tracking the Evolution of Member Roles in Practice Communities with Integrated Knowledge Graphs and Mining the Path to Building the Identity of Entrepreneurial Mentors

Member-role evolution and entrepreneurial mentor identification in Practice Communities have the following problems: high difficulty in dynamic tracking, insufficient integration of multi-dimensional features and implicit evolutionary path. Therefore, this paper proposes a two-layer analysis framework based on Knowledge Graph (KG): Firstly, a multi-dimensional knowledge graph based on member interaction, knowledge contribution and social network is constructed and member-role-behavior triplets are defined by entity extraction and relation mining; Secondly, a kind of Temporal Graph Convolutional Network (TGCN) is designed to capture trajectory of role evolution and PageRank algorithm is used to quantify node influence and discover the evolutionary path from peripheral participant to core mentor; Finally, a path mining algorithm is used to extract key transformation node and behavioral feature sequence. Experiments are based on three years of data from an Practice Communities, constructing a knowledge graph containing member nodes and 15293 relation edges. The average node degree is 10.74, the network density is 0.0038, and the clustering coefficient reaches 0.52, indicating that the community exhibits obvious smallworld characteristics. The results showed that the role classification module divided members into five roles: latent observers, casual participants, active contributors, domain experts, and entrepreneurial mentors, with a classification accuracy of 87.3%. Evolutionary trajectory analysis tracked the complete growth process of 217 members who eventually became entrepreneurial mentors, experiencing an average of 4.2 role transitions.