A Hybrid Recommendation System: A Review

• Published in: IEEE access Vol. 12; pp. 157107 - 157126
• Main Authors: Chaudhari, Anagha, Hitham Seddig, A. A., Sarlan, Aliza, Raut, Roshani
• Format: Journal Article
• Language: English
• Published: IEEE 2024
• Subjects: Accuracy; Big Data; Business; Collaborative filtering; concept diagrams; Concept drift; content-based filtering; Deep learning; Engines; Federated learning; Filtering; Information retrieval; Machine learning; Recommendation engines; Recommender systems
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2024.3480693

Over the years, recommendation engines (REs) have steadily increased in popularity, providing significant advantages by matching available items to users' specific interests. As more users, products, and rating information are introduced into the system, the relationship between users and recommended products changes, leading to a phenomenon known as concept drift (CD), which degrades the system's accuracy. Deep learning (DL), a subset of machine learning methodologies, involves learning over multiple layers of information processing phases and aids in addressing the CD problem. Additionally, federated recommender systems (FedRecSys) provide privacy for user data within a decentralised framework. This paper proposes a novel approach that hybridises DL methods with federated learning to detect and adapt to the concept drift issue in E-commerce-based REs. It also focuses on the benefits of collaborative RE and provides a systematic literature review of hybrid federated DL models to solve the challenges posed by concept drift for accurate recommendations. The results can be evaluated using performance measures such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Squared Log Error (MSLE), among others.