Abstract:

Higher education institutions (HEIs) worldwide are constantly under pressure to direct their students toward desirable labor-market requirements. During this process, HEIs must transform demand into learning outcomes while considering available resources. Curriculum creation is typically dependent on the experiences of academic personnel and, as such, might be regarded as subjective. The research described in this paper intends to demonstrate how machine learning methods based on data from HEI information systems can aid in curriculum creation. The authors also argue that data-driven reasoning can provide a more unbiased approach.

The authors used data from the information system of the Faculty of Information Technologies, University "Džemal Bijedić" in Mostar, Bosnia and Herzegovina. The data set comprised grades for students enrolled from 2017 to 2021. The goal was to see if using Principal Component Analysis (PCA) and Factor Analysis to determine subject grouping based on student grades was possible. The authors hypothesized that subjects would be grouped according to the grouping of learning outcomes if the assessment criteria were consistent with the learning outcomes.

The results reveal that PCA did not produce meaningful results, but factor analysis with the number of factors matching the number of learning outcome groups did. Data-supported grouping based on prior students' performance resulted in meaningful course grouping corresponding to learning outcome grouping. As a result, the authors argue that machine learning techniques can support the curriculum verification process. Furthermore, the authors claim that it is conceivable to use data-driven inference to calculate percentages expressing the significance of a learning outcome to a subject. The fact that when building a new curriculum, HEIs largely use the old modules (and learning outcomes), upgraded with a limited number of the new learning outcomes, demonstrates that the inference based on the previously collected data is informative and useful.

Keywords:

Curriculum development, machine learning, higher education.