Investigating Artificial Intelligence Algorithms to Predict College Students’ Academic Performance: A Systematic Mapping Study

The aim of this study is to analyse Artificial Intelligence (AI) algorithms in order to predict and identify Higher Education Institutions’ (HEI) students’ academic performance. A systematic mapping study (SMS) was carried out to investigate the limitations that researchers face when using AI algorithms to predict student academic performance and to determine which combinations of variables-algorithm yield the best results. A set of 43 studies was selected to be analysed. We found that the most commonly used variables to predict students’ academic performance can be grouped into socioeconomic (gender, age, and professional position), previous academic performance (grades, GPA, frequency, and scores at entrance exams), internet activity (use of LMS systems, e.g., Moodle or Google Classroom, and use of social media), and psychological and health (quality of sleep, eating habits, social life, academic or professional workload). The results show that the most common algorithm with the best evaluations to predict academic performance is the Random Forest at the time of publishing this study, the most common limitation faced by researchers is related to few available data, and the most common data used in these algorithms is related to previous academic performance.