Research Area:  Machine Learning
The tremendous expansion of information available on the web voraciously bombards users, leaving them unable to make decisions and having no way of stepping back to process it all. Recommender systems have emerged in this context as a solution to assist users by providing them with choices of appropriate and relevant items according to their preferences and interests. However, despite their success in many fields and application domains, they still suffer from the main limitation, known as the sparsity problem. The latter refers to the situation where insufficient transactional and feedback data are available for inferring specific user’s similarities, which affects the accuracy and performance of the recommender system. This paper provides a systematic literature review to investigate, analyze, and discuss the existing relevant contributions and efforts that use new concepts and tools to alleviate the sparsity issues. We have investigated the contributed similarity measures and have uncovered proposed approaches in different types of recommender systems. We have also identified the types of side information more commonly employed by recommender systems. Furthermore, we have examined the criteria that should be valued to enhance recommendation accuracy on sparse data. Each selected article was evaluated for its ability to mitigate the sparsity impediment. Our findings emphasize and accentuate the importance of sparsity in recommender systems and provide researchers and practitioners with insights on proposed solutions and their limitations, which contributes to the development of more powerful systems that can significantly solve the sparsity hurdle and thus enhance further the accuracy and efficiency of recommendations.
Author(s) Name:  Nouhaila Idrissi & Ahmed Zellou
Journal name:  Social Network Analysis and Mining
Publisher name:  Springer
Volume Information:  volume 10, Article number: 15 (2020)
Paper Link:   https://link.springer.com/article/10.1007/s13278-020-0626-2