Predicting Student’s Performance with Machine Learning: Challenges, Opportunities, and Future Directions

This paper describes and summarizes an analysis based on several students’ outcomes obtained at different moments throughout the semester in a curricular unit (CU) with topics related to Statistical Methods (SM). Evaluating student performance benefits both teachers and students: it helps students improve, and teachers adapt their teaching methods by identifying areas of struggle and predicting progress. With this goal in mind, machine learning and statistical techniques were applied to data acquired from a CU of the second year in an engineering program: Industrial Electronics and Computers Engineering (IECE) gathered over the past three academic years. The data included records of 299 students who completed several tasks: an individual assessment and at least six practical exercises developed in groups throughout seven weeks of the semester, defining a set of features as input. In the current study, machine learning (ML) algorithms will be used to predict individual students’ outcomes in the AS topic when performing different activities individually or in groups during the CU. This paper identifies the subjects and topics in SM that students fail or show more difficulties, and by identifying the most challenging topics, teachers can take proactive measures to enhance the quality of education and better prepare students for a rapidly evolving job market.