Machine Learning Applications for Early and Real-Time Warning Systems in Education

During the last decade, research has shown the accuracy and robustness of machine-learning methods for modelling nonlinear and complex relationships among several variables and for different prediction and classification purposes in the educational field. In this chapter we will present the main highlights coming from several studies that we carried out to predict a wide range of educational outcomes across different domains (reading readiness, mathematics, language, writing, general academic performance, and complex problem solving) using artificial neural networks. They have encompassed all educational levels from primary school to higher education, in different developed and developing countries. Furthermore, this approach was also useful to model key educational outcomes in academic trajectories such as retention, degree completion, and online education satisfaction in higher education. The main objective is to provide a comprehensive, integrated, systematic, and evidence-based framework for the application of machine learning systems in the educational field for improved prediction of performance, better understanding of factors influencing such performance, and improved policy decision-making regarding these kinds of problems. In addition, we will analyze and discuss the main specific patterns found in all these educational outcomes.