Scholarlink

<p>Recommendation systems (RSs) aim to provide personalized content to users by analyzing their interactions with various items. Context-aware recommendation systems (CARSs) extend this by considering contextual factors such as time and location, which strongly influence user decisions. However, incorporating such information often introduces complexity and scalability challenges. In this paper, we propose <b>CARA</b> (<b>C</b>ontext-<b>A</b>ware <b>R</b>ecommendation using Hierarchical <b>A</b>ttention and Positional Encoding Based on Multiple Item Attributes), a novel model that integrates hierarchical attention with enhanced positional encoding to capture both item order and attribute-aware relationships. Unlike prior approaches, CARA jointly models short-term and long-term preferences, producing more interpretable and context-sensitive recommendations. We evaluate CARA on two real-world benchmark datasets, MovieLens-100K and Yelp, and the results show consistent improvements. Specifically, CARA achieves improvements of 4% and 10% on MovieLens-100K, and 6% and 15% on Yelp, for Recall@20 and NDCG@20 respectively, compared to state-of-the-art baselines.</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

Context-aware recommendation using hierarchical attention and positional encoding based on multiple item attributes (CARA)

  • 期刊论文

  • 发表日期：

    2025年9月22日

查看全文
  • Hadise Vaghari,
  • Mehdi Hosseinzadeh Aghdam,
  • Hojjat Emami

摘要

Recommendation systems (RSs) aim to provide personalized content to users by analyzing their interactions with various items. Context-aware recommendation systems (CARSs) extend this by considering contextual factors such as time and location, which strongly influence user decisions. However, incorporating such information often introduces complexity and scalability challenges. In this paper, we propose CARA (Context-Aware Recommendation using Hierarchical Attention and Positional Encoding Based on Multiple Item Attributes), a novel model that integrates hierarchical attention with enhanced positional encoding to capture both item order and attribute-aware relationships. Unlike prior approaches, CARA jointly models short-term and long-term preferences, producing more interpretable and context-sensitive recommendations. We evaluate CARA on two real-world benchmark datasets, MovieLens-100K and Yelp, and the results show consistent improvements. Specifically, CARA achieves improvements of 4% and 10% on MovieLens-100K, and 6% and 15% on Yelp, for Recall@20 and NDCG@20 respectively, compared to state-of-the-art baselines.