Scholarlink

An investigation was undertaken to assess the effectiveness of item-based collaborative filtering (IBCF) within recommendation systems, particularly in scenarios where user preferences are significantly influenced by item attributes. The Pearson Correlation served as the chosen similarity metric in the User-Item Interaction Matrix. The efficiency of the recommender engine was adversely affected by a notable matrix sparsity of 93.7%, a computation based on the MovieLens 100K dataset. To mitigate this challenge, the study implemented the Singular Value Decomposition Matrix Factorization technique. After sparsity reduction and missing value imputation within the MovieLens 100K dataset, recommendations were regenerated. Notably, the proposed model demonstrated superior performance compared to the prior sparsity-based model. The efficiency increased from 75% pre-sparsity reduction to a noteworthy 79% following the implementation of the SVD algorithm.

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

From Sparsity to Accuracy: Enhancing Recommendation Systems with Item-Based Collaborative Filtering and SVD Matrix Factorization

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Arkajit Banerjee,
  • Bam Bahadur Sinha

摘要

An investigation was undertaken to assess the effectiveness of item-based collaborative filtering (IBCF) within recommendation systems, particularly in scenarios where user preferences are significantly influenced by item attributes. The Pearson Correlation served as the chosen similarity metric in the User-Item Interaction Matrix. The efficiency of the recommender engine was adversely affected by a notable matrix sparsity of 93.7%, a computation based on the MovieLens 100K dataset. To mitigate this challenge, the study implemented the Singular Value Decomposition Matrix Factorization technique. After sparsity reduction and missing value imputation within the MovieLens 100K dataset, recommendations were regenerated. Notably, the proposed model demonstrated superior performance compared to the prior sparsity-based model. The efficiency increased from 75% pre-sparsity reduction to a noteworthy 79% following the implementation of the SVD algorithm.