|
A Machine Learning Model for Investigating the Causes Influencing Student Dropout and Predicting Student Performance
|
Kiarash Zohori1    , Marjan Fallah2 |
1- Department of Toxicology and Pharmacology, Faculty of Pharmacy, Ayatollah Amoli Branch, Islamic Azad University, Amol, Iran , arieltehrani77@gmail.com
2- Department of Toxicology and Pharmacology, Faculty of Pharmacy, Ayatollah Amoli Branch, Islamic Azad University, Amol, Iran |
|
|
Abstract: (301 Views) |
Introduction: Early identification of students at risk of dropping out is essential for providing timely support, improving retention rates, and promoting academic success. Machine learning offers a powerful approach to analyze patterns in student behavior, allowing universities to predict dropouts and implement preventive interventions.
Methods: A hybrid model combining Particle Swarm Optimization and Extreme Gradient Boosting algorithms was developed to classify at-risk students. Additional data analysis techniques were applied to examine the dataset and extract meaningful insights.
Results: The proposed model achieved an accuracy of 98.12%, outperforming alternative models. Academic variables were identified as the most influential factors in dropout decisions. Students who failed to pay tuition on time were at substantially higher risk of dropping out (87.05%) and had a lower probability of graduation (4.95%), compared to those who paid on time, who exhibited a dropout rate of 25.21% and a graduation rate of 55.13%. Dropout rates also increased with age for both genders, although women consistently showed lower dropout rates across all age groups.
Conclusion: Machine learning-based prediction of dropout risk enables institutions to target interventions more effectively, optimizing resource allocation and supporting students in completing their education |
|
| Keywords: Machine learning, Metaheuristics, Student dropout, Decision-making |
|
|
|
Type of Article: Research Article |
Subject:
Health Management
Received: 2025/03/21 | Revised: 2025/11/25 | Accepted: 2025/09/15 | ePublished: 2025/10/7
|
|
|
|
|
|
|
| Send email to the article author |
|
|
