From Prediction to Practice: Interpretable Machine Learning for Proactive Student Retention Within Student Information Systems

Although several machine learning models are reported in the literature for student dropout prediction, only a few of them have been used in institutional Student Information Systems (SIS), in part because they need data or tools that are not available in the SIS. The purpose of this research is to put forward a deployment-oriented approach, mediated by only SIS-native data elements. The dataset is University of California, Irvine (UCI) Predict Students’ Dropout and Academic Success which contains the academic performance of students from the first and second semesters of their university studies, and the prediction of the students’ dropout is formulated as a binary classification problem; two models (Logistic Regression (LR) and Random Forest (RF)) are trained using the same data and tested with an 80/20 split. The accuracy for both models was 0.82, at-risk F1-score 0.71, and recall of 0.63, with receiver operating characteristic area under the curve (ROC-AUC) of 0.86 (LR) and 0.85 (RF). Results of the ablation study indicate that the F1-score can be improved to 0.79 when using all 36 of the dataset attributes while the evaluation of the age-group indicates that the false positive rate for students of age 25 or above is higher. The study also outlines a four-layer SIS integration architecture—data ingestion, risk prediction, advisor notification, and outcome tracking—which allows the LR model to both generate the risk score and give an explanation to the advisor, as well as a governance model for who is the owner, who has access to the predictions, how students are notified, and how long the risk scores and explanations are retained. The architecture is not in production yet and a staged evaluation plan is specified.

Authors

Institutions

Publication Details

Journal
Journal of Engineering Research and Sciences
Published
2026-11-17
DOI
https://doi.org/10.55708/js0511003
Primary Topic
Online Learning and Analytics
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

From Prediction to Practice: Interpretable Machine Learning for Proactive Student Retention Within Student Information Systems

Subrahmanya Bala Kiran Bhogaraju
Journal of Engineering Research and Sciences
Online Learning and Analytics
article

From Prediction to Practice: Interpretable Machine Learning for Proactive Student Retention Within Student Information Systems

Subrahmanya Bala Kiran Bhogaraju
article en

Abstract

Although several machine learning models are reported in the literature for student dropout prediction, only a few of them have been used in institutional Student Information Systems (SIS), in part because they need data or tools that are not available in the SIS. The purpose of this research is to put forward a deployment-oriented approach, mediated by only SIS-native data elements. The dataset is University of California, Irvine (UCI) Predict Students’ Dropout and Academic Success which contains the academic performance of students from the first and second semesters of their university studies, and the prediction of the students’ dropout is formulated as a binary classification problem; two models (Logistic Regression (LR) and Random Forest (RF)) are trained using the same data and tested with an 80/20 split. The accuracy for both models was 0.82, at-risk F1-score 0.71, and recall of 0.63, with receiver operating characteristic area under the curve (ROC-AUC) of 0.86 (LR) and 0.85 (RF). Results of the ablation study indicate that the F1-score can be improved to 0.79 when using all 36 of the dataset attributes while the evaluation of the age-group indicates that the false positive rate for students of age 25 or above is higher. The study also outlines a four-layer SIS integration architecture—data ingestion, risk prediction, advisor notification, and outcome tracking—which allows the LR model to both generate the risk score and give an explanation to the advisor, as well as a governance model for who is the owner, who has access to the predictions, how students are notified, and how long the risk scores and explanations are retained. The architecture is not in production yet and a staged evaluation plan is specified.

Journal of Engineering Research and SciencesVol. 5(11)
University of Denver (US)
Online Learning and Analytics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon