Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

ACADPro: XAI-student procrastination classification in academia using optuna optimized machine learning models.

V Jalaja Jayalakshmi1, M Punithavalli2

  • 1Department of Computer Applications, Bharathiar University, Coimbatore, India, 641046. vjalaja79@gmail.com.

Scientific Reports
|July 9, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Stress emotion classification using optimized convolutional neural network for online transfer learning dataset.

Computer methods in biomechanics and biomedical engineering·2022
See all related articles

This study introduces a reliability-driven framework to optimize machine learning models for predicting academic risk. The approach enhances decision-making by balancing predictive accuracy with reliability, outperforming single-classifier methods.

Area of Science:

  • Educational Psychology
  • Machine Learning
  • Data Science

Background:

  • Academic procrastination significantly impacts student achievement.
  • Existing methods for academic decision-making lack data-driven reliability.
  • Identifying academically vulnerable students requires robust predictive models.

Purpose of the Study:

  • To develop a novel, reliability-driven hyperparameter optimization framework for academic risk prediction.
  • To integrate discriminative capacity, predictive reliability, and falsified negative risks into a composite optimization problem.
  • To create an 'engineered academic risk' prediction target using behavioral and academic indicators.

Main Methods:

  • A composite optimization problem was defined, incorporating ROC-AUC, bootstrap stability, and a risk-sensitive penalty.
Keywords:
Academic procrastinationGradientBoostLogistic RegressionOptuna optimizationRandom ForestXAI

Related Experiment Videos

  • Preprocessing and feature selection were applied to prevent data leakage and circularity.
  • Hyperparameter optimization used Optuna with stratified five-fold cross-validation for robustness and reproducibility.

    • Main Results:

    • All tested classifiers demonstrated high predictive potential.
    • The Random Forest model achieved the highest composite score (0.9852), with strong ROC AUC (0.9766), accuracy (92.15%), and F1-score (0.9321).
    • Statistical significance tests confirmed the effectiveness of the reliability-focused optimization strategy over single-classifier approaches.

    Conclusions:

    • The proposed reliability-driven framework enables data-driven academic decisions by enhancing predictive model performance.
    • The engineered academic risk target and composite optimization effectively identify vulnerable students while minimizing false alarms.
    • The framework ensures statistical robustness and reproducibility, offering a more reliable alternative to traditional methods.