Ilmu Komputer & AI editorial
Open AccessOA2024
Predicting University Student Graduation Using Academic Performance and Machine Learning: A Systematic Literature Review
This systematic literature review analyzes 70 journal articles (2018โ2023) on machine learning-based prediction of university student graduation, identifying Support Vector Machine and Random Forest as the most common algorithms and Learning Management System and Student Information System data as primary sources.
Lidya R Pelima; Yuda Sukmana; Y. Rosmansyahยท IEEE Accessยท 2024ยท DOI 10.1109/ACCESS.2024.3361479
The core problem
Predicting university student graduation is beneficial for both students and institutions. Students can make informed decisions about academic and career paths, while institutions can proactively identify at-risk students and offer tailored support. Machine learning (ML) has gained attention for this task, with some studies reporting accuracy rates as high as 90%. However, existing systematic literature reviews (SLRs) have limitations: they often omit discussion of predictive models and algorithms, lack coverage of ML algorithms, use small databases, select keywords that miss relevant synonyms, and provide limited transparency in data collection. This study addresses these gaps by conducting a thorough SLR of ML-based student graduation prediction models from 70 journal articles published between 2018 and 2023. The review examines algorithms implemented, academic performance data used, model effectiveness, and challenges and potential benefits. The goal is to advance smart learning environments by providing a nuanced and comprehensive review.
Innovation
The authors conducted a systematic literature review following a structured process. They searched multiple databases using keywords related to student graduation prediction, machine learning, and academic performance, including synonyms to ensure comprehensive coverage. The search was limited to journal articles published from 2018 to 2023. After screening, 70 relevant articles were selected for analysis. Data extraction focused on: (1) machine learning algorithms used, (2) types of academic performance data (e.g., from Learning Management Systems (LMS) and Student Information Systems (SIS)), (3) model performance metrics, and (4) reported challenges and benefits. The review aimed to identify trends, gaps, and best practices in the field.
Introduction
Predicting university student graduation is beneficial for both students and institutions. Students can make informed decisions about academic and career paths, while institutions can proactively identify at-risk students and offer tailored support. Machine learning (ML) has gained attention for this task, with some studies reporting accuracy rates as high as 90%. However, existing systematic literature reviews (SLRs) have limitations: they often omit discussion of predictive models and algorithms, lack coverage of ML algorithms, use small databases, select keywords that miss relevant synonyms, and provide limited transparency in data collection. This study addresses these gaps by conducting a thorough SLR of ML-based student graduation prediction models from 70 journal articles published between 2018 and 2023. The review examines algorithms implemented, academic performance data used, model effectiveness, and challenges and potential benefits. The goal is to advance smart learning environments by providing a nuanced and comprehensive review.
Why it matters
The findings indicate that while ML can effectively predict student graduation, there are limitations in the existing literature. The predominance of SVM and RF suggests these algorithms are robust for this task, but other algorithms may be underutilized. The reliance on LMS and SIS data highlights the importance of institutional data infrastructure. However, the lack of detailed reporting on model development and evaluation hinders reproducibility and comparison. The review also notes that challenges such as data privacy, ethical considerations, and the need for interpretable models remain. The potential benefits include early intervention, personalized support, and improved graduation rates. The study contributes to smart learning environments by providing a comprehensive overview and identifying gaps for future research. Future work should explore a wider range of algorithms, standardize reporting, and address ethical concerns.
Who should read this
Pembaca Research Digest dan praktisi rumpun ini.
Opening member contentโฆ