A1 Vertaisarvioitu alkuperäisartikkeli tieteellisessä lehdessä
Unleashing the Power of Predictive Analytics to Identify At-Risk Students in Computer Science
Tekijät: Bin Qushem Umar, Oyelere Solomon Sunday, Akçapınar Gökhan, Kaliisa Rogers, Laakso Mikko-Jussi
Kustantaja: SPRINGER
Julkaisuvuosi: 2023
Journal: Technology Knowledge and Learning
Tietokannassa oleva lehden nimi: TECHNOLOGY KNOWLEDGE AND LEARNING
Lehden akronyymi: TECHNOL KNOWL LEARN
Sivujen määrä: 16
ISSN: 2211-1662
eISSN: 2211-1670
DOI: https://doi.org/10.1007/s10758-023-09674-6
Verkko-osoite: https://link.springer.com/article/10.1007/s10758-023-09674-6
Tiivistelmä
Predicting academic performance for students majoring in computer science has long been a significant field of research in computing education. Previous studies described that accurate prediction of students' early-stage performance could identify low-performing students and take corrective action to improve performance. Besides, adopting machine learning algorithms with predictive analytics has proven possible and meaningful. The traditional approach of looking after students without uncovering the root causes of poor performance has shifted dramatically into improving the quality of the educational processes of students, teachers, and stakeholders. Thus, this study employed predictive analytics to develop an early warning prediction model using computing science degree performance data at a public institution. Predictive models based on our data analysis revealed that low, medium, and high-performing students could be predicted with an accuracy of 88% using only the grades of the courses they took in the second year. Moreover, 96% accuracy was achieved when all course grades were used in predictive models. The courses that are important in determining the overall performance of the students were also analyzed. By employing a multi-method approach, utilizing a large dataset spanning four academic years, and including a diverse sample of 430 students, our study offers a robust foundation to researchers, designers, and computer science educators for understanding and predicting student performance. The enhanced generalizability and implications for educational practice position our study as a valuable contribution to the field, paving the way for further advancements in predictive analytics.
Predicting academic performance for students majoring in computer science has long been a significant field of research in computing education. Previous studies described that accurate prediction of students' early-stage performance could identify low-performing students and take corrective action to improve performance. Besides, adopting machine learning algorithms with predictive analytics has proven possible and meaningful. The traditional approach of looking after students without uncovering the root causes of poor performance has shifted dramatically into improving the quality of the educational processes of students, teachers, and stakeholders. Thus, this study employed predictive analytics to develop an early warning prediction model using computing science degree performance data at a public institution. Predictive models based on our data analysis revealed that low, medium, and high-performing students could be predicted with an accuracy of 88% using only the grades of the courses they took in the second year. Moreover, 96% accuracy was achieved when all course grades were used in predictive models. The courses that are important in determining the overall performance of the students were also analyzed. By employing a multi-method approach, utilizing a large dataset spanning four academic years, and including a diverse sample of 430 students, our study offers a robust foundation to researchers, designers, and computer science educators for understanding and predicting student performance. The enhanced generalizability and implications for educational practice position our study as a valuable contribution to the field, paving the way for further advancements in predictive analytics.
Turun yliopiston Tutkimustietojärjestelmän portaali