Akademik Veri Yönetim Sistemi
Arabian Journal for Science and Engineering, 2026 (SCI-Expanded, Scopus)
Human resources analytics stands out as a critical area for enhancing employee engagement and satisfaction. However, the privacy and security of data related to employee behavior pose a significant barrier in such analyses and applications. To address issues related to data privacy and sharing, the federated learning architecture offers a new approach in the development of employee engagement prediction models. Federated learning preserves data privacy by enabling the processing of data located in different organizations without centralization. Additionally, sharing models contributes to reducing network traffic and energy consumption while also decreasing the carbon footprint. In this study, a federated learning-based relational classification architecture was developed on the IBM HR Analytics dataset to improve employee engagement prediction. During the model training, the duCBA (Data Unaware Classification Based on Association) algorithm was revised, and a new federated learning aggregation algorithms called FedCBA (Federated Classification Based on Association) was used. The experimental results show that the model achieves %87 accuracy and proves the effectiveness of the proposed method. Findings reveal that employees who work overtime, have low environmental and job satisfaction, and have short tenure are likelier to leave their jobs. In contrast, employees with good work–life balance and high satisfaction levels tend to stay. These findings suggest that the proposed method protects data confidentiality and provides tangible insights to improve employee engagement.