Scholarlink

<p>It is a popular neighborhood-based recommendation method called Collaborative Filtering (CF) employs similarity measures to provide personalized recommendations to users. Traditional similarity measures often rely solely on historical ratings, overlooking the reality that user tastes evolve with time. To deal with this limitation, this study introduces PWaHy-CAGNet, a novel Parallel Walrus Hyperbolic Convolutional Attention Graph Network for enhancing collaborative filtering recommendations. The proposed approach integrates Grid-Constrained Data Cleansing for preprocessing, multiple correlation metrics (Cosine Correlation Coefficient (COS), Pearson Correlation Coefficient (PCC), Gower’s Coefficient), and advanced time-aware similarity refinement using Exponential and Power Decay functions. By incorporating hyperbolic convolution and attention mechanisms, PWaHy-CAGNet effectively captures complex user-item interactions. Experimental results Show that Gower's Coefficient is applicable to the MovieLens-100k dataset with exponential decay achieves superior performance, validating the model’s efficacy in personalized recommendations. The investigation shows that With a lower RMSE of 0.6836 and a higher accuracy of 99%, the DSHS-ConAtNet model respectively.</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

RETRACTED ARTICLE: Advancing time-based recommendations in collaborative filtering: evaluating performance with parallel Walrus hyperbolic convolutional attention graph networks

  • 期刊论文

  • 发表日期：

    2025年6月6日

查看全文
  • Rajiv Kumar Nath,
  • Tanvir Ahmad

摘要

It is a popular neighborhood-based recommendation method called Collaborative Filtering (CF) employs similarity measures to provide personalized recommendations to users. Traditional similarity measures often rely solely on historical ratings, overlooking the reality that user tastes evolve with time. To deal with this limitation, this study introduces PWaHy-CAGNet, a novel Parallel Walrus Hyperbolic Convolutional Attention Graph Network for enhancing collaborative filtering recommendations. The proposed approach integrates Grid-Constrained Data Cleansing for preprocessing, multiple correlation metrics (Cosine Correlation Coefficient (COS), Pearson Correlation Coefficient (PCC), Gower’s Coefficient), and advanced time-aware similarity refinement using Exponential and Power Decay functions. By incorporating hyperbolic convolution and attention mechanisms, PWaHy-CAGNet effectively captures complex user-item interactions. Experimental results Show that Gower's Coefficient is applicable to the MovieLens-100k dataset with exponential decay achieves superior performance, validating the model’s efficacy in personalized recommendations. The investigation shows that With a lower RMSE of 0.6836 and a higher accuracy of 99%, the DSHS-ConAtNet model respectively.