Scholarlink

<p>Recently, adaptive learning platforms have attracted attention for their ability to provide personalized learning experiences. However, existing recommendation methods face limitations in effectively dealing with users’ learning behaviors and diverse learning resources. This paper proposes a hierarchical graph-based multi-type learning resource recommendation model. First, it creates a hierarchical structure graph to explore associations between different learning resources and learns these features using graph neural networks. Secondly, it proposes a knowledge tracing module that integrates the positive and negative learning behaviors. Then it studies users’ short-term preferences from interactive sequences through RNN with knowledge gating, and users’ long-term preferences with an attention mechanism. It further adaptively fuses the long-term and short-term learning preferences based on context weighting. Finally, it utilizes a feature pyramid network to integrate users’ learning preferences at different levels. The output of the upper layer is used to guide the lower layer. The model summarizes the recommendation results of each layer and generates a recommendation list of multiple types of learning resources. Extensive experiments on several public datasets demonstrate that the proposed model effectively improves the performance of recommender systems in several application scenarios.</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

Learning resource recommendation models based on learning behaviors and hierarchical structure graph

  • 期刊论文

  • 发表日期：

    2025年10月4日

查看全文
  • Lihua Bai,
  • Heng Wang,
  • Zhi Zhang,
  • Qing Xie,
  • Mengzi Tang

摘要

Recently, adaptive learning platforms have attracted attention for their ability to provide personalized learning experiences. However, existing recommendation methods face limitations in effectively dealing with users’ learning behaviors and diverse learning resources. This paper proposes a hierarchical graph-based multi-type learning resource recommendation model. First, it creates a hierarchical structure graph to explore associations between different learning resources and learns these features using graph neural networks. Secondly, it proposes a knowledge tracing module that integrates the positive and negative learning behaviors. Then it studies users’ short-term preferences from interactive sequences through RNN with knowledge gating, and users’ long-term preferences with an attention mechanism. It further adaptively fuses the long-term and short-term learning preferences based on context weighting. Finally, it utilizes a feature pyramid network to integrate users’ learning preferences at different levels. The output of the upper layer is used to guide the lower layer. The model summarizes the recommendation results of each layer and generates a recommendation list of multiple types of learning resources. Extensive experiments on several public datasets demonstrate that the proposed model effectively improves the performance of recommender systems in several application scenarios.