Scholarlink

<p>Collaborative filtering systems leverage user preference data to generate accurate recommendations for new items. A critical challenge in this process is the presence of "gray sheep" users—individuals with atypical preferences who introduce noise and reduce recommendation accuracy for "white" users, whose preferences align with the general population. While kMeans is a widely used clustering algorithm for clustering users, its performance is highly sensitive to the initial selection of cluster centers, often leading to convergence on suboptimal local optima. To address this limitation, this paper introduces a novel hybrid approach named HybridFireflyEkMeans. This model combines the EkMeans algorithm (an optimized version of kMeans that includes advanced variants such as kMeans++, PkMeans++, and MkMeans++) with the Firefly metaheuristic algorithm. The Firefly algorithm is employed to select optimal initial cluster centers, thereby ensuring a more robust and efficient clustering process that avoids local optima. Initial cluster centers for the firefly population are seeded with solutions generated by the EkMeans variants, which provides a strong starting point for the optimization process. Experimental results on two standard datasets, MovieLens and FilmTrust, demonstrate that our proposed HybridFireflyEkMeans model significantly improves clustering quality and user separation. This, in turn, enhances recommendation accuracy for white users, confirming the efficacy of our approach in addressing a key challenge in collaborative filtering systems<b>.</b></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

HybridFireflyEkMeans: a novel approach to enhance collaborative filtering through optimized user clustering and gray sheep user separation

  • 期刊论文

  • 发表日期：

    2026年6月3日

查看全文
  • Juan Yu,
  • Rahim Rashidi

摘要

Collaborative filtering systems leverage user preference data to generate accurate recommendations for new items. A critical challenge in this process is the presence of "gray sheep" users—individuals with atypical preferences who introduce noise and reduce recommendation accuracy for "white" users, whose preferences align with the general population. While kMeans is a widely used clustering algorithm for clustering users, its performance is highly sensitive to the initial selection of cluster centers, often leading to convergence on suboptimal local optima. To address this limitation, this paper introduces a novel hybrid approach named HybridFireflyEkMeans. This model combines the EkMeans algorithm (an optimized version of kMeans that includes advanced variants such as kMeans++, PkMeans++, and MkMeans++) with the Firefly metaheuristic algorithm. The Firefly algorithm is employed to select optimal initial cluster centers, thereby ensuring a more robust and efficient clustering process that avoids local optima. Initial cluster centers for the firefly population are seeded with solutions generated by the EkMeans variants, which provides a strong starting point for the optimization process. Experimental results on two standard datasets, MovieLens and FilmTrust, demonstrate that our proposed HybridFireflyEkMeans model significantly improves clustering quality and user separation. This, in turn, enhances recommendation accuracy for white users, confirming the efficacy of our approach in addressing a key challenge in collaborative filtering systems.