Evaluating the Long Tail in Recommendation System: A Systematic Review of Approaches, Datasets and Metrics

A recommendation system (RS) leverages machine learning by analyzing user behavior, and suggesting relevant products based on the user's preferences. Long-tail items, which were once leading products in their niche, became harder to find and newer items are heavily promoted to users, long-tail items, which can boost customer engagement and ensure that long-tail items remain visible. In this paper, we have provided extensive efforts to conduct a systematic review of the long-tail recommendation system, based on PRISMA 2020 guidelines. Studies published between 2012 and 2024 were identified, through a detailed search in ACM Digital Library, Science Direct, SpringerLink, IEEE Xplore, and Google Scholar. We conducted a detailed investigation into a long-tail recommendation system which focused on finding different categories, datasets, and evaluation metrics. This literature review provides an extensive overview of the selection of datasets, different categories of long-tail recommendation, and evaluation criteria for the researchers and individuals who are new to the domain of long-tail recommendation systems.