Scholarlink

Traditional recommender systems face limitations such as the cold-start problem in Collaborative Filtering (CF) and limited contextual understanding in Content-Based Filtering (CBF). This paper proposes a hybrid recommendation framework integrating CF using Singular Value Decomposition (SVD), enriched CBF leveraging TF-IDF metadata vectors, and semantic similarity via Sentence Transformer embeddings indexed with FAISS (Facebook AI Similarity Search). The hybrid model balances CF predictions and semantic similarity scores through a tunable parameter. Evaluated on the MovieLens 100k dataset, results demonstrate that this integrated approach significantly improves Precision@K and Recall@K compared to standalone CF or CBF models. Moreover, we discuss explicit applications of our hybrid recommender system in sustainable development scenarios, such as optimizing resource allocation, reducing waste through precise recommendations, and promoting environmentally friendly choices.

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 Recommender Systems Leveraging Collaborative Filtering, Semantic Embeddings, and FAISS for Sustainable Development Applications

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Charaf Hamidi,
  • Mohamed Badiy,
  • Abdelaaziz Hessane,
  • Hind Aiouej,
  • Hicham Tribak,
  • Salma Gaou

摘要

Traditional recommender systems face limitations such as the cold-start problem in Collaborative Filtering (CF) and limited contextual understanding in Content-Based Filtering (CBF). This paper proposes a hybrid recommendation framework integrating CF using Singular Value Decomposition (SVD), enriched CBF leveraging TF-IDF metadata vectors, and semantic similarity via Sentence Transformer embeddings indexed with FAISS (Facebook AI Similarity Search). The hybrid model balances CF predictions and semantic similarity scores through a tunable parameter. Evaluated on the MovieLens 100k dataset, results demonstrate that this integrated approach significantly improves Precision@K and Recall@K compared to standalone CF or CBF models. Moreover, we discuss explicit applications of our hybrid recommender system in sustainable development scenarios, such as optimizing resource allocation, reducing waste through precise recommendations, and promoting environmentally friendly choices.