Scholarlink

<p>In high-sparsity scenarios, collaborative filtering is susceptible to the impact of data sparsity, which leads to deviations in neighbor identification and further results in low recommendation accuracy and diversity. To address this issue, this paper proposes a multi-granular rough k-means clustering model. Firstly, the model adopts the mean method and the min-max similarity method to optimize the initial clustering centers of k-means, thereby alleviating the local optimality problem. Secondly, it accurately identifies neighbors in sparse data by constructing the lower and upper approximation sets of rough k-means grains. Finally, by dynamically adjusting the rough coefficient and applying it to collaborative filtering recommendation, multi-granularity switching is achieved, the global rough granule set is mined to capture multi-level interests, and an interpretable multi-granularity recommendation list is provided for target users. Experimental results demonstrate that on both the MovieLens and Beauty datasets, the proposed model not only significantly mitigates the impact of data sparsity but also improves the accuracy and diversity of recommendations.</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

Multi-granularity rough k-means clustering in the optimization of collaborative filtering algorithms

  • 期刊论文

  • 发表日期：

    2026年4月10日

查看全文
  • Ziru Wang,
  • Jusheng Mi,
  • Ziyun Li

摘要

In high-sparsity scenarios, collaborative filtering is susceptible to the impact of data sparsity, which leads to deviations in neighbor identification and further results in low recommendation accuracy and diversity. To address this issue, this paper proposes a multi-granular rough k-means clustering model. Firstly, the model adopts the mean method and the min-max similarity method to optimize the initial clustering centers of k-means, thereby alleviating the local optimality problem. Secondly, it accurately identifies neighbors in sparse data by constructing the lower and upper approximation sets of rough k-means grains. Finally, by dynamically adjusting the rough coefficient and applying it to collaborative filtering recommendation, multi-granularity switching is achieved, the global rough granule set is mined to capture multi-level interests, and an interpretable multi-granularity recommendation list is provided for target users. Experimental results demonstrate that on both the MovieLens and Beauty datasets, the proposed model not only significantly mitigates the impact of data sparsity but also improves the accuracy and diversity of recommendations.