Silalahi, Sondius Matogu Budiman and Indrajit, Richardus Eko and Mantoro, Teddy Development of a Prediction Model for Academic Performance Decline Using Two-Layer Ensemble Progressive Learning. IEEE.
Sondius.pdf
Download (498kB)
Abstract
Educational data mining (EDM) has emerged as
a critical field for leveraging academic data to improve decision
making in schools and support early interventions for at-risk
students. Despite the availability of rich academic data, school
counseling services often remain reactive, with interventions
occurring only after performance declines, creating a gap for
proactive early-warning systems. To address this, a Two-Layer
Ensemble Progressive Learning model is proposed: the first
layer integrates Support Vector Machine(SVM), K-Nearest
Neighbors (KNN), and Random Forest (RF) through weighted
voting, while the second layer fuses inter-semester predictions to
capture progression dynamics. Adaptive curriculum features
covering curriculum-alignment behavior and knowledge
domain grouping are engineered to be readily replicable across
schools, relying only on routinely collected curriculum and
assessment records. On the aggregated evaluation set (N = 240),
we obtain micro-averaged Accuracy = 0.929, Precision = 0.723,
Recall = 0.895, F1 = 0.800, and ROC AUC = 0.902. Under 10
fold grouped, time-aware cross-validation, the ensemble
achieves macro-averaged Recall = 0.912 and F1 = 0.866. The
confusion matrix (TP = 34, FN = 4, FP = 13, TN = 189) highlights
strong recall and low false negatives, which are critical for early
interventions. This study makes three key contributions: a
progressive ensemble that leverages inter-semester dynamics;
adaptive curriculum features that are readily replicable across
schools; and a practical implementation flow that enables school
counselors to move from reactive to proactive intervention.
Keywords— Educational Data Mining, Early Warning
System, Ensemble Progressive Learning, Machine Learning,
Replicable Features, Student Performance Prediction.
| Item Type: | Article |
|---|---|
| Subjects: | -|- SUBJEK PRADITA -|- > Fakultas Sains dan Teknologi > Magister Teknologi Informasi |
| Divisions: | Fakultas Sains dan Teknologi > Magister Teknologi Informasi |
| Depositing User: | Pradita Librarian |
| Date Deposited: | 15 Oct 2026 03:02 |
| Last Modified: | 15 Oct 2026 03:02 |
| URI: | https://repository.pradita.ac.id/id/eprint/744 |
