Predicting academic performance of university students at high risk of dropping out of programming courses: a literature review

Abstract This paper presents a literature review on predictive models of academic performance and their role in identifying university students at high risk of dropping out from programming courses. The PRISMA methodology was employed, defining the phases of identification, screening, selection, and inclusion. Various databases were consulted, and 60 articles were selected for qualitative analysis. Factors such as low academic grades, insufficient time, unmet goals, stress, anxiety, and socioeconomic conditions are identified as key determinants of dropout. The combined use of Learning Management Systems (LMS) and Early Warning Systems (EWS) integrated with machine learning algorithms has been shown to reduce dropout rates in these courses by approximately 14%. Classification algorithms and neural networks are effective tools for identifying students at high risk of dropping out. The primary variables utilized in these algorithms relate to academic, psychological, and socioeconomic factors. Future research should focus on implementing these models within LMS and EWS tools to develop early and personalized interventions.