Answers are grounded in this paper and cite their sources.
It is crucial for educational institutions to understand their students' academic levels of achievement in order to measure their academic progress and give students appropriate academic supports (Ahmed et al., 2025).There are different methods for evaluating students' academic levels of achievement.Some methods include rubrics, such as Bloom's Taxonomy; the CIPP (context, input, process, and product) evaluation model; and regression and fuzzy logic, as well as the rapid increase of the use of ML and DL.Thus far, research studies have not come to an agreement on which category of models is most appropriate to use in different educational settings.Many comparative studies have looked at algorithms that belong in the same category.This paper provides a systematic comparative review of models used to evaluate academic performance and classifies them in four main categories, which include: traditional/rubricbased, statistical/soft computing, classical ML and DL.Models are assessed in five dimensions, which include, predictive accuracy, interpretability, data requirements, computability and scalability.Recent studies point to Random Forest and AdaBoost as having a high level of predictive accuracy; however, these models may not be as interpretable as more traditional and fuzzy-logic based models.Traditional and fuzzy-logic based models may be preferred if aligning with educational goals is prioritized over achieving high level of predictive accuracy.In general, no category of models has been found to be the best across all contexts.The model selection should be based on the purpose of the evaluation, whether it be formative or for predictive outcome, and the data readiness of the institution.Potential research areas include AI, hybrid fuzzy-logic based ML and explanatory ML.
No takes yet. Share an insight, caveat, or question.
Okpobo et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: