Scholarlink

<p>This study presents an AI-powered talent chain management framework for supervised talent mobility prediction, with a particular focus on next-occupation classification from career histories. Given an individual’s chronologically ordered ESCO occupation-code sequence and job-title text sequence, the model predicts the ESCO occupation label corresponding to the next career transition. Existing transition-based and neural sequence models often rely on local occupational statistics or generic sequential representations, making it difficult to jointly capture structured occupational taxonomy information, job-title semantics, temporal career dynamics, and uncertainty in heterogeneous career records. To address these limitations, we propose an adaptive talent dynamics planner built upon a multi-agent modeling perspective. The framework integrates a constraint-driven workforce optimizer, an agent-based collaboration forecaster, and an uncertainty-aware mobility evaluator to represent talent states, occupational compatibility, collaboration-related dependencies, and risk-aware decision refinement. In the prediction model, ESCO-code embeddings and job-title text representations are fused into transition-state representations, temporal order encoding captures direction-sensitive career dynamics, and uncertainty-aware refinement regularizes noisy or ambiguous occupational transitions through perturbation consistency. Experiments are conducted on KARRIEREWEGE and the DECORTE career histories dataset under a unified supervised evaluation protocol. The proposed model achieves the best performance on both datasets, obtaining 0.241 Accuracy@1, 0.562 Recall@10, 0.351 MRR@10, and 0.421 NDCG@10 on KARRIEREWEGE, and 0.216 Accuracy@1, 0.510 Recall@10, 0.316 MRR@10, and 0.380 NDCG@10 on DECORTE. Compared with the strongest baseline CAREER, it improves NDCG@10 by 2.7 and 2.9 percentage points on the two datasets while requiring fewer parameters and FLOPs. These results demonstrate that the proposed framework improves exact next-occupation prediction, top-10 ranking quality, robustness, and deployment feasibility for ESCO-based talent mobility analysis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

AI-powered talent chain management with multi-agent systems for industry and innovation growth

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月30日

查看全文
  • Rongfu Wang,
  • Xiufen Zeng,
  • Fuchao Li,
  • Bin Wang,
  • Juan Zhang

摘要

This study presents an AI-powered talent chain management framework for supervised talent mobility prediction, with a particular focus on next-occupation classification from career histories. Given an individual’s chronologically ordered ESCO occupation-code sequence and job-title text sequence, the model predicts the ESCO occupation label corresponding to the next career transition. Existing transition-based and neural sequence models often rely on local occupational statistics or generic sequential representations, making it difficult to jointly capture structured occupational taxonomy information, job-title semantics, temporal career dynamics, and uncertainty in heterogeneous career records. To address these limitations, we propose an adaptive talent dynamics planner built upon a multi-agent modeling perspective. The framework integrates a constraint-driven workforce optimizer, an agent-based collaboration forecaster, and an uncertainty-aware mobility evaluator to represent talent states, occupational compatibility, collaboration-related dependencies, and risk-aware decision refinement. In the prediction model, ESCO-code embeddings and job-title text representations are fused into transition-state representations, temporal order encoding captures direction-sensitive career dynamics, and uncertainty-aware refinement regularizes noisy or ambiguous occupational transitions through perturbation consistency. Experiments are conducted on KARRIEREWEGE and the DECORTE career histories dataset under a unified supervised evaluation protocol. The proposed model achieves the best performance on both datasets, obtaining 0.241 Accuracy@1, 0.562 Recall@10, 0.351 MRR@10, and 0.421 NDCG@10 on KARRIEREWEGE, and 0.216 Accuracy@1, 0.510 Recall@10, 0.316 MRR@10, and 0.380 NDCG@10 on DECORTE. Compared with the strongest baseline CAREER, it improves NDCG@10 by 2.7 and 2.9 percentage points on the two datasets while requiring fewer parameters and FLOPs. These results demonstrate that the proposed framework improves exact next-occupation prediction, top-10 ranking quality, robustness, and deployment feasibility for ESCO-based talent mobility analysis.