Graph-Based Analytical Approach to Identifying Substitute Human Resources: Integrating Individual Capabilities and Group Dynamics

In today’s volatile business environment, securing a sustainable competitive advantage hinges on retaining and effectively managing talent. While talent turnover is inevitable, strategic internal human resource (HR) transfers offer a solution to prevent talent outflow and supplement skill gaps. However, previous models for identifying internal substitutes often focus solely on individual work capabilities, neglecting the critical role of group interactions and collaborative structure. Drawing on social network theory, transactive memory systems, and person–group fit, this study proposes a graph-based analytical approach that models the organization as a complex system. Our methodology provides a holistic framework that integrates both (1) individual capabilities and (2) group-level characteristics (e.g., work-relationship networks and cluster-level similarity) to identify the most suitable substitutes. At the macroscopic level, we use an inductive graph neural network (GraphSAGE) to learn node embeddings from a work relationship network constructed from process event logs and to quantify group-level similarity. At the microscopic level, we compute dynamic collaboration intensity, frequency, and task similarity between employees over time. To validate the approach, we develop four simulation scenarios using an enriched incident management process event log and implement them in a SimPy-based simulator, benchmarking against an existing method that considers only individual factors. Across all scenarios, the proposed dual-factor model significantly outperforms the baseline in terms of efficiency, accuracy, and suitability. This research provides a practical, validated algorithm that supports evidence-based workforce management and more effective internal talent allocation.