Calibration-aware decision-support policy for student dropout early warning

Wijiyanto Wijiyanto, Aris Marjuni, Amad Zainul Fanani, Ruri Suko Basuki, Nurmalitasari Nurmalitasari

Abstract


Given the limited intervention capacity, dropout early warning systems must be operationalized as auditable decision-support policies rather than standalone predictive models. This study developed and evaluated a calibrated, recall-constrained early-warning policy using data from 329 students at a private university in Central Java, Indonesia, with academic, socioeconomic, and perceptual/relational predictors, to prioritize dropout risk for intervention triage. The workflow integrated: i) predictive modeling evaluated under nested cross-validation, ii) probability calibration to obtain reliable risk scores, and iii) constraint-based decision thresholding (minimum recall target), combined with risk band stratification (low/medium/high/very high), to map probabilities to actionable service levels. Performance was assessed across imbalance-sensitive utility area under the precision-recall curve (AUPRC), discrimination area under the receiver operating characteristic curve (ROC-AUC), and probability reliability (Brier score), complemented by significance testing and ablation/importance analyses for policy interpretability. Results indicated that calibration and recall-based threshold tuning improved decision usability while preserving competitive predictive performance. The primary contribution was an explicit decision-support layer (calibration, threshold tuning, and risk bands) with empirical validation and factor-driven guidance for implementation. Accordingly, the resulting calibration-aware decision-support policy provided an auditable basis for translating student dropout risk estimates into prioritized intervention actions under limited institutional capacity.

Keywords


Decision support; Early warning system; Probability calibration; Risk stratification; Threshold optimization

Full Text:

PDF


DOI: https://doi.org/10.11591/eei.v15i6.12300

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Bulletin of EEI Statistics

Bulletin of Electrical Engineering and Informatics (BEEI)
ISSN: 2089-3191 , e-ISSN: 2302-9285
This journal is published by the Institute of Advanced Engineering and Science (IAES) .