Abstract:

The rapid growth in making digital educational environments has generated huge numbers of datasets, enabling the application of data-driven decision-making or prediction using Machine Learning (ML) models with high precision. However, the adoption of these ML models is often hindered by their "black-box" nature to the educators as they output predictions without context or reasoning. Therefore, along with the predictability power of ML models, there is a need for transparent, interpretable feedback through the lens of Explainable AI (XAI) that educators can use for supporting improved student learning. Considering this fact, this study aims to examine student demographic and engagement data to analyze student performance prediction together with the interpretable output for knowing proper reasoning and context. This approach can further explore the critical intersection of predictability produced by ML models. So, in this study, a set of ML algorithms (including Gradient Boosting, Random Forest, K-Nearest Neighbors, Support Vector Machines, Naive Bayes and Decision Tree) are applied to a student dataset to build a number of prediction models for student performance analysis. After that, by integrating XAI frameworks such as SHAP (SHapley Additive exPlanations), this study demonstrates how complex model outputs can translate into actionable pedagogical insights for supporting enhanced learning in education. The findings suggest that prioritizing interpretability power of XAI in education not only boosts trust among stakeholders but also empowers educators to design data-driven interventions that address the specific behavioral and academic drivers of student success through personalized intervention.

Keywords:

Machine Learning, Explainable AI, Student Performance, Educational Data Mining, Learning Analytics, Interpretability, Education.