Academic institutions dedicate a substantial effort to ensure the academic success of their students. At the course level, teachers recommend learning paths for students to guarantee the achievement of their learning outcomes. In terms of performance, these kinds of approaches are deemed more effective than others based uniquely on performing a collection of independent activities. However, there is no systematic means to validate if following the learning path is effective, nor to assess whether and to what extent students adhere to these recommendations. This paper introduces a novel technique for modelling recommended learning paths, including not only an evaluation of path utility, but also a quantitative measure of student adherence thereto using process mining, and more precisely, log skeletons. Following an event abstraction process regarding real student-recorded activity, a scoping process is employed to retain the trajectories that adhere to the prescribed learning path. The method based on process mining is translated into practice by considering an online university mathematics course. Results confirm the applicability of the method and, in this case, reveal that adhering to the suggested path correlates positively with final grades. Few students strictly follow the prescribed learning path, although the vast majority support it. The method can be easily applied to overcome several challenges associated with enhancing academic performance from the learning analytics perspective.

High dropout rates constitute a major concern for higher education institutions, due to their economic and academic impact. The problem is particularly relevant for institutions offering online courses, where withdrawal ratios are reported to be higher. Both the impact and these high rates motivate the implementation of interventions oriented to reduce course withdrawal and overall institutional dropout. In this paper, we address the identification of populations of learners at risk of withdrawing from higher education online courses. This identification is oriented to design interventions and is carried out using survival analysis. We demonstrate that the method’s longitudinal approach is particularly suited for this purpose and provides a clear view of risk differences among learner populations. Additionally, the method quantifies the impact of underlying factors, either alone or in combination. Our practical implementation used an open dataset provided by The Open University. It includes data from more than 30,000 students enrolled in different courses. We conclude that low-income students and those who report a disability comprise risk groups and are thus feasible intervention targets. The survival curves also reveal differences among courses and show the detrimental effect of early dropout on low-income students, worsened throughout the course for disabled students. Intervention strategies are proposed as a result of these findings. Extending the entire refund period and giving greater academic support to students who report disability are two proposed strategies for reducing course withdrawal.