A Mobile Graph Neural Network Framework for Organizational Structure Modeling and Employee Performance and Turnover Analysis

Abstract

With the widespread adoption of mobile technologies, organizational interaction data has become increasingly heterogeneous and dynamic. Existing human resource analytics models struggle to simultaneously address data privacy protection, dynamic organizational relationship modeling, and mobile deployment requirements, limiting their ability to accurately capture the influence of organizational structures on employee performance and turnover. To address these challenges, this paper proposes an end-to-end, mobile-oriented interactive human resource analytics framework based on graph neural networks (GNNs). The framework enables a full-cycle process from dynamic organizational interaction data collection and modeling to analysis and managerial decision support. Specifically, it innovatively constructs dynamic heterogeneous graphs to achieve unified quantification and dynamic modeling of multidimensional relationships within organizations. A spatiotemporalaware dynamic graph neural network (DGNN) is designed to handle temporal evolution and distinguish multiple relational types. Furthermore, a privacy-preserving federated graph learning framework is introduced to accommodate distributed mobile data storage scenarios. Finally, a lightweight model and mobile-side interactive inference scheme are proposed to overcome deployment of bottlenecks for complex models on mobile devices. Experimental results demonstrate that the proposed model significantly outperforms existing mainstream methods in employee performance prediction and turnover intention identification tasks, while meeting real-time inference and resource constraints on mobile platforms. This framework provides a reliable technical foundation for mobile human resource management.