Scholarlink

This paper presents a recommender algorithm integrating a multi-agent ensemble case-based reasoning (ESCBR-SMA), a Thompson sampling-based (TS) recommender system, and a Hawkes process. The final integrated algorithm is applied to improve the real-time adaptation of an Intelligent Tutoring System called AI-VT. We have compared the static recommendation algorithm (ESCBR-SMA with TS) and the dynamic recommendation algorithm (ESCBR-SMA, TS with the Hawkes process) by evaluating the knowledge acquisition evolution of each learner. The metrics used allow us to determine the stability of prediction and change in the probability distributions for each learner and each level of complexity. The results show that the integration between stochastic adaptation, the prediction with the case-based reasoning paradigm, and the Hawkes process allows reinforcement of knowledge as well as a more realistic estimation of the recommendation for each case independently.

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

Ensemble Stacking Case-Based Reasoning and a Stochastic Recommender Algorithm with the Hawkes Process Applied to ITS AI-VT

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Daniel Soto-Forero,
  • Marie-Laure Betbeder,
  • Julien Henriet

摘要

This paper presents a recommender algorithm integrating a multi-agent ensemble case-based reasoning (ESCBR-SMA), a Thompson sampling-based (TS) recommender system, and a Hawkes process. The final integrated algorithm is applied to improve the real-time adaptation of an Intelligent Tutoring System called AI-VT. We have compared the static recommendation algorithm (ESCBR-SMA with TS) and the dynamic recommendation algorithm (ESCBR-SMA, TS with the Hawkes process) by evaluating the knowledge acquisition evolution of each learner. The metrics used allow us to determine the stability of prediction and change in the probability distributions for each learner and each level of complexity. The results show that the integration between stochastic adaptation, the prediction with the case-based reasoning paradigm, and the Hawkes process allows reinforcement of knowledge as well as a more realistic estimation of the recommendation for each case independently.