Explainable Artificial Intelligence (XAI) is increasingly applied to student-performance prediction to improve the transparency and interpretability of machine-learning decisions. Local Interpretable Model-Agnostic Explanations (LIME) provides model-agnostic local explanations; however, its effectiveness depends on the configuration of parameters governing neighbourhood construction, perturbation sampling and feature selection. Conventional LIME relies largely on default or manually selected configurations, which may produce explanations with inconsistent fidelity, interpretability and computational efficiency when applied to educational data.  This study developed an Osprey-Optimised LIME (OOA-LIME) framework that employs the Osprey Optimisation Algorithm (OOA) to automatically search for effective LIME configurations. The framework was evaluated using 10,000 anonymised student records from a distance-learning programme. Standard LIME and OOA-LIME were evaluated across dataset sizes of 2,000–10,000 records using fidelity, interpretability, comprehensibility, trustworthiness, coverage and explanation-generation time.  OOA-LIME consistently outperformed Standard LIME across all five explanation-quality measures. Fidelity increased from 0.850–0.890 for OOA-LIME compared with 0.780–0.812 for Standard LIME, while interpretability increased from 0.800–0.836 compared with 0.710–0.738. Comprehensibility, trustworthiness and coverage similarly improved. Explanation-generation time decreased from 12.01–12.98 seconds for Standard LIME to 7.58–8.09 seconds for OOA-LIME. The findings indicate that metaheuristic optimisation can improve both the explanatory quality and computational efficiency of LIME for educational prediction. OOA-LIME provides an optimisation-based approach for improving LIME configurations and offers a practical framework for more effective explainable student-performance prediction in distance-learning environments.

Keywords: Explainable Artificial Intelligence (XAI), Student Performance Prediction, Local Interpretable Model-Agnostic Explanations (LIME), Osprey Optimisation Algorithm (OOA), Model Interpretability, Educational Data Mining.