“Application of machine learning to curriculum design analysis,”

Abstract

Curriculum re-design typically involves a review of the relative strengths and weaknesses of an academic programme particularly with respect to the suitability pre- and co-requisites of courses. Courses that are identified as requisite sometimes provide little conceptual scaffolding, with the result being a curriculum design whose sequence does not correlate with the progression routes adopted by students. An application of Identification Trees is proposed here to determine the sensitivity of a course to the anticipated requisites independently of human bias and based entirely on the available data. Among the various Machine Learning approaches considered, Identification Trees is one of the few which generates results can be interpreted in terms of a priority of courses that determine course outcome sensitivity to requisites. A secondary approach is also through cross-correlation using the Pearson r-coefficient. Both approaches were applied to academic programmes in the School of Engineering and a predictive accuracy of 72% was demonstrated even at long correlation capacity of 32%. Overall, the method gave reliable results and is hereby proposed as a potentially useful analysis technique in the design of curricula.