A Fuzzy Clustering-Based Personalized Recommendation Algorithm for Online Shopping

Recommendation systems are essential for delivering personalized content and enhancing user engagement on digital platforms, particularly as the volume of users and product offerings grows continuously. However, with the rapid expansion of online content, recommendation systems encounter significant challenges. Maintaining real-time performance, achieving high-quality recommendations, and mitigating data sparsity and user homogeneity are among the key issues that traditional methods, especially collaborative filtering, often fail to address. These challenges highlight the importance of developing advanced recommendation technologies capable of improving both recommendation accuracy and diversity. In response to these challenges, this research enhances the Fuzzy C-Means algorithm by integrating Entropy Weighting and Slope Correction, enabling more adaptive clustering and improved recommendation accuracy under sparse data conditions. By leveraging the flexible partitioning capability of the fuzzy C-means clustering algorithm, the proposed method effectively reduces data dimensionality and improves recommendation quality in sparse data environments. Additionally, the approach’s adaptive partitioning feature enables the system to dynamically adjust to diverse user patterns, enhancing its real-time processing capabilities and recommendation precision. Our experimental results demonstrate a 12% reduction in Mean Absolute Error (MAE) and a 15% increase in recommendation diversity compared to standard collaborative filtering techniques, marking significant advancements in both accuracy and diversity. Furthermore, the method performed optimally within a 15–25 neighbor count range, showing its effectiveness across various datasets. This improvement reflects the system's capability to provide more nuanced and varied recommendations, thereby enriching user experience. These findings underscore the potential of fuzzy clustering techniques to support recommendation systems in handling complex, large-scale environments, achieving a 20% improvement in processing speed. By successfully addressing the limitations of collaborative filtering, this study contributes a robust solution for more accurate, diverse, and real-time recommendation generation, which is crucial for platforms with extensive user bases and diverse content libraries.

Supervisor: Professor Sa'Ed Abed Convener: 
Dr. Asma Abdulrahman AlSumait
Examination Committee: Dr. Abdullah Alshaibani