Citation

Singla, Sanjay and Mittal, Harish Kumar and Kang, Sandeep Singh and Singla, Priti and Senthilpari, Chinnaiyan and Azmeera, Rahul Lightweight Explainable AI Framework for Early Academic Risk Prediction and Sustainable Student Support in Engineering Education. Social Science Forum, 10 (7). pp. 655-675. ISSN 03523608

Text 16.pdf - Published Version Restricted to Repository staff only Download (207kB)

Abstract

Early identification of academically at-risk students is important for improving retention, academic performance and timely institutional support. This study proposes a lightweight explainable artificial intelligence framework that reframes heterogeneous academic outcome prediction tasks as three intervention-oriented risk levels: High Risk, Moderate Risk and Low Risk. The framework is evaluated on a public higher-education dropout dataset and a Bachelor of Technology (B.Tech) semester-grade dataset. Five tuned machine learning models are compared under a class-imbalance-aware protocol using macro F1-score as the primary metric. On the primary dataset, histogram-based gradient boosting achieved the best performance, with a macro F1-score of 0.7112 and accuracy of 0.7751, while Logistic Regression remained a competitive transparent baseline. On the B.Tech dataset, Extra Trees achieved a macro F1-score of 0.9708. Feature-group ablation, permutation importance, Shapley additive explanations and odds-ratio analysis consistently show that curricular-unit approval, semester progress and fee status dominate prediction, while socio-demographic variables have limited standalone value. Threshold-sensitivity and multi-seed testing support robustness. Student-level risk evidence cards translate predictions into mentor-readable intervention cues, positioning the framework as accountable decision support rather than automated student labelling.

Item Type:	Article
Uncontrolled Keywords:	Educational data mining, explainable artificial intelligence
Subjects:	L Education > L Education (General)
Divisions:	Faculty of Artificial Intelligence & Engineering (FAIE)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	04 Sep 2026 07:13
Last Modified:	04 Sep 2026 07:13
URII:	http://shdl.mmu.edu.my/id/eprint/16735

Downloads

Edit (login required)