Scholarlink

<p>In the rapidly advancing environment of online platforms, recommender systems play a pivotal role in adapting content to user preferences. However, achieving a balance between personalization, fairness, and discovery remains a significant challenge for conventional recommendation methods. This paper introduces FairRecMOEA/D, a novel multi-objective optimization approach combining decomposition-based evolutionary algorithms (MOEA/D) with collaborative and demographic filtering. FairRecMOEA/D simultaneously optimizes recall, F1 score, novelty, and fairness. Evaluations using diverse MovieLens datasets demonstrate its superior ability to balance these conflicting objectives, outperforming traditional models. Our results underscore the transformative potential of multi-objective optimization for enhancing user engagement and satisfaction, setting a new benchmark for future recommender systems.</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-objective Optimization for Personalized and Fairness Recommender Systems: A Hybrid Approach

  • 期刊论文

  • 发表日期：

    2025年7月31日

查看全文
  • Fatima Ezzahra Zaizi,
  • Sara Qassimi,
  • Said Rakrak

摘要

In the rapidly advancing environment of online platforms, recommender systems play a pivotal role in adapting content to user preferences. However, achieving a balance between personalization, fairness, and discovery remains a significant challenge for conventional recommendation methods. This paper introduces FairRecMOEA/D, a novel multi-objective optimization approach combining decomposition-based evolutionary algorithms (MOEA/D) with collaborative and demographic filtering. FairRecMOEA/D simultaneously optimizes recall, F1 score, novelty, and fairness. Evaluations using diverse MovieLens datasets demonstrate its superior ability to balance these conflicting objectives, outperforming traditional models. Our results underscore the transformative potential of multi-objective optimization for enhancing user engagement and satisfaction, setting a new benchmark for future recommender systems.