In recent years, the use of online teaching platforms has grown rapidly, offering students flexibility and access to distance learning. However, the absence of in-person interactions between educators and learners can make it difficult to identify students in difficulty and provide them with appropriate support. On the other hand, thanks to technological advances, e-learning platforms can collect and analyze large amounts of data on students, such as their performance, their interactions with content and their learning habits. In this article, we present a review of recent literature on models for identifying students with learning difficulties on e-learning platforms in an academic environment, in order to provide them with targeted activities at the right time to improve their learning. Four questions have therefore been selected: How can we identify students with online learning difficulties? What are the parameters for this identification? What Machine Learning (ML) techniques are currently used to detect students with online learning difficulties? And what are their limitations? How can we identify students with online learning difficulties without using machine learning and their limitations? To answer these questions, Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) technique literature review serves as the foundation for our methodology.