Scholarlink

<p>Technological advancement has led to online learning platforms, and with these, recommending relevant courses to the users have caused challenges such as data sparsity and cold start problem. As a result of these challenges, traditional recommendation systems fail thus providing less relevant and personalized courses recommendation. Thus, the given paper proposes a hybrid course recommendation system which utilizes singular value decomposition (SVD), non-negative matrix factorization (NMF) and Attention-based Neural Network. Because of this combination, the model can produce extremely personalized recommendations based on user’s interactions with the application and inherent features of the courses. Dynamic Profile Matching is used to addresses the cold start problem by dynamically adjusting to new users or objects, resulting in useful suggestions even with limited interaction data. Mean absolute error (MAE) and root mean square error (RMSE) are used to assess the performance of the proposed system. The results show that recommendation accuracy has improved significantly, with the proposed hybrid model’s RMSE of 0.6998 and MAE of 0.6737 being 17% better than those of the conventional models.</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

Hybrid course recommendation system using SVD, NMF and attention-based neural network

  • 期刊论文

  • 发表日期：

    2025年2月20日

查看全文
  • Mohd Mustafeez ul Haque,
  • Bonthu Kotaiah,
  • Jameel Ahamed

摘要

Technological advancement has led to online learning platforms, and with these, recommending relevant courses to the users have caused challenges such as data sparsity and cold start problem. As a result of these challenges, traditional recommendation systems fail thus providing less relevant and personalized courses recommendation. Thus, the given paper proposes a hybrid course recommendation system which utilizes singular value decomposition (SVD), non-negative matrix factorization (NMF) and Attention-based Neural Network. Because of this combination, the model can produce extremely personalized recommendations based on user’s interactions with the application and inherent features of the courses. Dynamic Profile Matching is used to addresses the cold start problem by dynamically adjusting to new users or objects, resulting in useful suggestions even with limited interaction data. Mean absolute error (MAE) and root mean square error (RMSE) are used to assess the performance of the proposed system. The results show that recommendation accuracy has improved significantly, with the proposed hybrid model’s RMSE of 0.6998 and MAE of 0.6737 being 17% better than those of the conventional models.