One clear example of research that uses educational data mining (EDM) as a source of evidence is the study “Educational data mining: prediction of students’ academic performance using machine learning algorithms” published in Smart Learning Environments. This research applied EDM techniques to a dataset of 1,854 undergraduate students’ records from a Turkish university to predict final exam grades based on midterm scores and other academic attributes. By employing several machine‑learning algorithms—such as random forests, support vector machines, logistic regression, and k‑nearest neighbours—the researchers built predictive models capable of classifying final performance with around 70–75% accuracy. The study highlights how EDM can discover hidden relationships in educational data and contribute to early prediction of at‑risk students, supporting decision‑making and intervention strategies in higher education. Springer Link

This kind of research shows several things that educational data mining can tell us:

Predictions of academic outcomes: EDM models can estimate the likelihood that a student will succeed or struggle based on prior performance and interaction patterns. In the cited study, grades from earlier assessments were strong predictors of final performance. Springer Link

Patterns in learning behavior: By analyzing large datasets, EDM can uncover relationships between variables that might not be obvious—such as correlations between participation data and achievement. Springer Link

Identification of at‑risk students: Predictive modelling can signal students who may need support early in a course, giving educators a chance to intervene. Springer Link

However, EDM also has limitations and things it cannot tell us on its own:

Why learning outcomes occur: EDM may indicate which variables correlate with success but doesn’t fully explain why those relationships exist or capture deeper qualitative context. For example, it can show that midterm grades predict final grades but not why some students struggle beyond numerical patterns. SpringerLink

Non‑quantifiable skills: Skills such as creativity, collaboration, and critical thinking are difficult to infer directly from typical dataset variables like scores or clicks. MDPI

Equity and socioeconomic nuance: Many EDM studies rely on academic and demographic data but may not fully reflect complex socioeconomic factors that influence learning, especially if those factors are not captured in the datasets. MDPI

In summary, research using educational data mining—such as the performance prediction study—demonstrates EDM’s ability to predict academic outcomes and reveal hidden patterns, offering valuable evidence for interventions and instructional decisions. At the same time, EDM cannot replace deeper qualitative understanding or capture aspects of learning that lie outside structured data.