Home // IARIA Congress 2023, The 2023 IARIA Annual Congress on Frontiers in Science, Technology, Services, and Applications // View article

Forecasting Failure Risk in Early Mathematics and Physical Science Courses in the Bachelor's Degree in Engineering

Authors:
Isaac Caicedo-Castro
Mario Macea-Anaya
Samir Castaño-Rivera

Keywords: machine learning; quantum machine learning; educational data mining; supervised learning; classification methods; failure forecasting.

Abstract:
In this research, we study the functional mapping between university admission test scores and the likelihood of failure in initial mathematics and physics courses for students embarking on a Bachelor's degree in Systems Engineering. We assume that the admission test assesses students' competence and proficiency in natural sciences and mathematics, essential prerequisites for success in the foundational courses of this Systems Engineering program. A deficiency in these subjects might result in failure, leading to dropouts or an extended degree completion timeline. We harnessed machine learning techniques to probe this issue, focusing on the landscape of Colombian universities, specifically analyzing the Systems Engineering program at the University of Córdoba. In this Colombian educational context, universities, including our case study institution, rely on the national standardized admission test known as Saber 11 to evaluate candidates for Bachelor's degree programs. By adopting machine learning methods to unveil underlying patterns that govern this functional mapping, we might proactively identify students at risk of struggling in the aforementioned courses based on their admission test scores. Early identification of these at-risk students opens the opportunity to preemptive measures, such as offering preparatory courses to fortify their prerequisites for success in these challenging subjects. Our research involved the examination of academic records from 56 anonymized students, using both 10-fold and 5-fold cross-validation. The outcomes from the 10-fold cross-validation reveal that the support vector machine method yields mean values of 71.33% for accuracy, 68.33% for precision, 60% for recall, and 62.05% for the harmonic mean (F1). Therefore, we conclude that this method outperforms the others studied in this work.

Pages: 177 to 187

Copyright: Copyright (c) IARIA, 2023

Publication date: November 13, 2023

Published in: conference

ISBN: 978-1-68558-089-6

Location: Valencia, Spain

Dates: from November 13, 2023 to November 17, 2023