Addressing sparse data challenges in recommendation systems: A systematic review of rating estimation using sparse rating data and profile enrichment techniques

Abstract

E-commerce recommendation systems enhance the user experience by providing customized suggestions tailored to user preferences. They analyze user interactions, such as ratings, to identify user preferences and recommend relevant items accordingly. The sparsity of user–item rating data poses a significant obstacle for Recommender Systems, making it difficult to model user preferences effectively. This issue is particularly evident in collaborative filtering techniques, where the accuracy of user/item similarity calculations and latent factor identification are compromised. Therefore, the sparsity adversely affects the accuracy, coverage, scalability, and transparency of the recommendations, posing significant challenges. Various approaches have been developed to estimate sparse values from the available ratings and improved user/item profiles with side information or sparse ratings to improve the estimation by addressing the four challenges. Despite extensive research in this area, there is a lack of comprehensive surveys that specifically explore estimation methods using sparse rating data and profile enrichment techniques focusing on overcoming challenges that occur due to sparsity such as reduced coverage, transparency, scalability, and accuracy. Understanding the effectiveness of these approaches, their impact on recommendations and future research directions is a crucial area of the literature. This study seeks to examine the individual effects of rating-based estimation and profile enrichment-based estimation methods in Collaborative Filtering recommender systems, to address challenges related to sparsity, identify research gaps, and suggest future research directions. It also provides readers with information on methodologies, available datasets with varying levels of sparsity, and ongoing research challenges in this field.