Scholarlink

<p>Group recommendation systems (GRS) are designed to generate personalized recommendations for groups of users with varying preferences. One of the key challenges in this domain lies in precisely identifying and forming user groups based on similarity in preferences, as group membership is often not predefined. Conventional similarity measures such as Pearson Correlation, Cosine Similarity, and Jaccard struggle to capture complex and nonlinear patterns in user behaviour, especially in sparse or noisy data environments. In this work, we propose a novel similarity computation method, HFSDSim: Hybrid Fuzzy Stability Deviation, specifically designed for effective user clustering in group recommendation systems. We also propose a novel GRS framework based on similarity. The HFSDSim measure integrates fuzzy logic with statistical and distance-based metrics, dynamically adjusting weights to better handle uncertainty and nonlinearity in user preferences. This similarity measure is used as the basis for forming user clusters, thereby improving intra-group cohesion and inter-group distinction. To evaluate its effectiveness, HFSDSim is compared against traditional and state-of-the-art similarity metrics including PCC (Pearson Correlation Coefficient), Cosine Similarity (CS), Adjusted Cosine Similarity (ACOS), Jaccard Similarity (JS), PIP (Proximity Impact Popularity), CPCC (Constrained Pearson Correlation Coefficient), PSS (Proximity Significance Singularity), NHSM (New Heuristic Similarity Model), and HSMD (Hybrid Difference-based Similarity Measure) on three benchmark datasets MovieLens, FilmTrust, and IMDB. Evaluation metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Squared Error (MSE), are used to assess the accuracy of predicted ratings within groups. The model achieves superior accuracy, with average reductions in MAE and MSE of up to 20–25% compared to the best competing methods. Experimental results demonstrate that HFSDSim consistently outperforms existing methods, highlighting its potential as an effective foundation for group formation in GRS.</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

A new hybrid similarity-based framework for effective group recommendation system

  • 期刊论文

  • 发表日期：

    2026年6月16日

查看全文
  • Shilpa Singhal,
  • Kunwar Pal

摘要

Group recommendation systems (GRS) are designed to generate personalized recommendations for groups of users with varying preferences. One of the key challenges in this domain lies in precisely identifying and forming user groups based on similarity in preferences, as group membership is often not predefined. Conventional similarity measures such as Pearson Correlation, Cosine Similarity, and Jaccard struggle to capture complex and nonlinear patterns in user behaviour, especially in sparse or noisy data environments. In this work, we propose a novel similarity computation method, HFSDSim: Hybrid Fuzzy Stability Deviation, specifically designed for effective user clustering in group recommendation systems. We also propose a novel GRS framework based on similarity. The HFSDSim measure integrates fuzzy logic with statistical and distance-based metrics, dynamically adjusting weights to better handle uncertainty and nonlinearity in user preferences. This similarity measure is used as the basis for forming user clusters, thereby improving intra-group cohesion and inter-group distinction. To evaluate its effectiveness, HFSDSim is compared against traditional and state-of-the-art similarity metrics including PCC (Pearson Correlation Coefficient), Cosine Similarity (CS), Adjusted Cosine Similarity (ACOS), Jaccard Similarity (JS), PIP (Proximity Impact Popularity), CPCC (Constrained Pearson Correlation Coefficient), PSS (Proximity Significance Singularity), NHSM (New Heuristic Similarity Model), and HSMD (Hybrid Difference-based Similarity Measure) on three benchmark datasets MovieLens, FilmTrust, and IMDB. Evaluation metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Squared Error (MSE), are used to assess the accuracy of predicted ratings within groups. The model achieves superior accuracy, with average reductions in MAE and MSE of up to 20–25% compared to the best competing methods. Experimental results demonstrate that HFSDSim consistently outperforms existing methods, highlighting its potential as an effective foundation for group formation in GRS.