Meta-model for Recommendation of Machine Learning Algorithm in Education

The increasing use of artificial intelligence (AI) and machine learning (ML) in education has created
the need for clear guidance on selecting appropriate
algorithms for specific educational tasks.
Educational datasets differ considerably in size,
structure, and pedagogical context, while machine
learning algorithms vary in complexity,
interpretability, and practical feasibility.
Consequently, educators and researchers-
particularly those without strong technical
backgrounds- rely on trial-and-error approaches,
which can result in inefficient or poorly justified
analytical choices. This paper proposes an
explainable meta-model for recommending machine
learning algorithms in educational settings. The
approach is grounded in meta-learning and focuses
on capturing relationships between dataset
characteristics and algorithm performance. Four
classification methods-Ridge Classifier, Decision
Tree Classifier, Multilayer Perceptron, and K-
Nearest Neighbors-were used to develop and
evaluate the meta-model across 15 educational
datasets. Model performance was assessed using
multiple metrics relevant to educational decision-
making, including accuracy, precision, recall, F1-
score, Cohen’s kappa, and computational time,
alongside a cross-validation strategy to ensure
robustness. The results highlight important trade-offs
between predictive accuracy, reliability,
interpretability, and efficiency. While more complex
models achieved higher accuracy, simpler and more
transparent approaches demonstrated more
balanced performance and greater practical
suitability for educational and policy-oriented
contexts. These findings reinforce the value of
context-aware algorithm selection rather than
reliance on single performance metrics. The
proposed meta-model offers practical support for
educators and policy makers by reducing
experimentation costs and enabling evidence-based
decision-making. By bridging algorithmic complexity
and educational needs, this research contributes a
generalizable and interpretable framework that
promotes trustworthy use of machine learning in
education