Article: Core talent loss early warning algorithm integrating temporal collaborative filtering and the Prophet model Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.66 pp.20 - 38 Abstract: In the knowledge economy, core talent loss threatens technological and trade secret leakage, hindering organisational growth. Current prediction methods often neglect temporal dynamics. This study addresses this by integrating temporal collaborative filtering (TCF) and the Prophet model into a novel early warning algorithm. TCF extracts dynamic temporal patterns in employee behaviour, while Prophet captures trend and seasonal features. A dual-view deep neural network fuses both approaches, supported by a dedicated data pipeline. Experimental results demonstrate that the fusion model achieves 85.2% test accuracy, surpassing TCF-only (76.1%) and Prophet-only (73.4%) models by 9.1 and 11.8 percentage points, respectively, with 82.7% recall and a 0.863 F1 score. Under a 6-month time window and 23,890 training samples, the model attains 85.7% accuracy, 81.5% coverage of key-period loss events, and 0.62-second response time, confirming its effectiveness for talent loss warning. Inderscience Publishers - linking academia, business and industry through research

Title: Core talent loss early warning algorithm integrating temporal collaborative filtering and the Prophet model

Authors: Hao Yu

Addresses: School of Management, Zhengzhou Shengda University, Zhengzhou, 451191, China

Abstract: In the knowledge economy, core talent loss threatens technological and trade secret leakage, hindering organisational growth. Current prediction methods often neglect temporal dynamics. This study addresses this by integrating temporal collaborative filtering (TCF) and the Prophet model into a novel early warning algorithm. TCF extracts dynamic temporal patterns in employee behaviour, while Prophet captures trend and seasonal features. A dual-view deep neural network fuses both approaches, supported by a dedicated data pipeline. Experimental results demonstrate that the fusion model achieves 85.2% test accuracy, surpassing TCF-only (76.1%) and Prophet-only (73.4%) models by 9.1 and 11.8 percentage points, respectively, with 82.7% recall and a 0.863 F1 score. Under a 6-month time window and 23,890 training samples, the model attains 85.7% accuracy, 81.5% coverage of key-period loss events, and 0.62-second response time, confirming its effectiveness for talent loss warning.

Keywords: talent loss early warning; temporal collaborative filtering; Prophet model; fusion algorithm; employee churn prediction; temporal forecasting; early warning system.

DOI: 10.1504/IJICT.2026.154124

International Journal of Information and Communication Technology, 2026 Vol.27 No.66, pp.20 - 38

Received: 18 Nov 2025
Accepted: 07 Jan 2026
Published online: 13 Jun 2026 *

Title: Core talent loss early warning algorithm integrating temporal collaborative filtering and the Prophet model

Keep up-to-date