Article: Dynamic employee turnover risk prediction model based on GraphSAGE-LSTM Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.66 pp.57 - 78 Abstract: Employee turnover prediction is crucial for corporate human resource strategies, yet traditional static models struggle to capture complex employee relationship networks and behavioural temporal dynamics. To address this, this paper proposes a graph sample and aggregation - long short-term memory fusion model that analyses employee collaboration relationships via graph neural networks while modelling historical behavioural sequences using long short-term memory networks, thereby enabling dynamic and precise turnover risk prediction. Experiments on the publicly available dataset demonstrate that the proposed model achieves an area under the receiver operating characteristic curve of 0.92, representing an improvement of approximately 0.07 over a single long short-term memory model. Its accuracy rate reaches 89.5%, surpassing traditional logistic regression by nearly 12%. This research not only validates the effectiveness of integrating graph structure and temporal information to enhance predictive performance but also provides reliable technical support for enterprises to implement early-stage risk intervention. Inderscience Publishers - linking academia, business and industry through research

Title: Dynamic employee turnover risk prediction model based on GraphSAGE-LSTM

Authors: Guangyu Li

Addresses: School of Management, The Open University of Shaanxi, Xi'an, 710000, China

Abstract: Employee turnover prediction is crucial for corporate human resource strategies, yet traditional static models struggle to capture complex employee relationship networks and behavioural temporal dynamics. To address this, this paper proposes a graph sample and aggregation - long short-term memory fusion model that analyses employee collaboration relationships via graph neural networks while modelling historical behavioural sequences using long short-term memory networks, thereby enabling dynamic and precise turnover risk prediction. Experiments on the publicly available dataset demonstrate that the proposed model achieves an area under the receiver operating characteristic curve of 0.92, representing an improvement of approximately 0.07 over a single long short-term memory model. Its accuracy rate reaches 89.5%, surpassing traditional logistic regression by nearly 12%. This research not only validates the effectiveness of integrating graph structure and temporal information to enhance predictive performance but also provides reliable technical support for enterprises to implement early-stage risk intervention.

Keywords: employee turnover risk; graph neural networks; GNNs; dynamic prediction.

DOI: 10.1504/IJICT.2026.154127

International Journal of Information and Communication Technology, 2026 Vol.27 No.66, pp.57 - 78

Received: 13 Feb 2026
Accepted: 21 Mar 2026
Published online: 13 Jun 2026 *

Title: Dynamic employee turnover risk prediction model based on GraphSAGE-LSTM

Keep up-to-date