In competitive sports and other role-based team activities, selecting a squad with strong mutual compatibility is critical for coordinated execution and overall performance. This paper proposes a compatibility-aware graph search framework to recommend optimal team compositions in distributed edge platforms where athlete profiles, training logs, and collaboration records are naturally scattered across clubs, devices, or data silos. We model athletes as nodes in a weighted graph, where edge weights encode tactical fit and historical synergy, and node attributes capture role capabilities and task requirements. Team construction is formulated as a Group Steiner Tree (GST) problem that seeks a minimum-cost connected subgraph covering at least one candidate from each required role group, while implicitly balancing individual competence and cross-member collaboration potential. To make GST practical for distributed settings, we introduce a search strategy with pruning and incremental expansion that reduces the combinatorial space and supports partial aggregation across sites. Experiments on simulated datasets demonstrate that the proposed method consistently finds more compatible and cost-effective team configurations than heuristic baselines, improving the quality of tactical lineup planning and training-group recommendation in distributed edge environment.
Li et al. (Thu,) studied this question.