Home // International Journal On Advances in Intelligent Systems, volume 17, numbers 1 and 2, 2024 // View article
Authors:
Isaac Caicedo-Castro
Keywords: Machine learning; educational data mining; classification algorithm
Abstract:
In this paper, we address the problem of predicting whether a student might fail a course before it starts, based on their academic history. This study is centered on predicting failure in the numerical methods course, which is part of the curriculum for the bachelor’s degree in systems engineering at the University of Co ́rdoba in Colombia. To tackle this problem, we adopt classification methods from supervised machine learning. To this end, we utilize a dataset initially collected in [1] and subsequently expanded in [2]. This dataset is used to fit and validate the machine learning methods employed in this study. Our work contributes to improving the quality of the forecasting task compared to prior research [1], [2]. This improvement has been achieved by modifying the vector representation of the student’s academic history, considering only the student’s performance in mathematics, as evidenced in the admission test called Saber 11 and prerequisite courses. The results of the experimental validation reveal that the method based on Gaussian processes with the Radial Basis Function achieves mean values of accuracy, precision, recall, and harmonic mean of 83%, 80.67%, 77%, and 76.70%, respectively. This method has outperformed the others studied in this work. Moreover, the prediction outcome of Gaussian processes is the probability that a given student will fail the course, which is convenient for designing an intervention plan to help them succeed. Therefore, the conclusion of this study is twofold. Firstly, Gaussian processes are the best choice to implement an intelligent system for the prediction task studied herein. Secondly, this study finds a clear correlation between the probability of succeeding in the numerical methods course and the student’s competencies in mathematics obtained before enrolling in this course. This suggests that good training in mathematics courses is required to succeed in the numerical methods course.
Pages: 25 to 37
Copyright: Copyright (c) to authors, 2024. Used with permission.
Publication date: June 30, 2024
Published in: journal
ISSN: 1942-2679