Search

This paper combines a theoretical-based model with a data-driven approach to develop an Early Warning System that detects students who are more likely to dropout. The model uses innovative multilevel statistical and machine learning methods. The paper demonstrates the validity of the approach by applying it to administrative data from a leading Italian university.

Study success in Higher Education is of primary importance in the European policy agenda. Yet, given the diverse educational landscape across countries and institutions, more coordinated action is needed to gain a more solid knowledge of the dropout phenomenon. This study aims to gain a better insight into students' dropout based on an integrated comparative study of two universities located in two different European countries: Politecnico di Milano (Italy) and Vrije Universiteit Amsterdam (the Netherlands). This research aims at assessing whether the factors affecting dropout are similar in the Italian and the Dutch contexts by testing the predictive capacity of ad-hoc models trained in other university-country settings at three different stages of the student's university journey: (i) enrolment, (ii) end of the first semester, and (iii) end of the first year. Results show that the predictive capacity of models is exchangeable across different contexts, and it improves dramatically once data on university performance becomes available. We find that the models trained in the Dutch context have a better ability to identify dropouts in the Italian context than the other way around. Models trained on Dutch data allow us to better understand the relationship between educational credits obtained, the most important variable across models, and students' dropout. This study contributes to creating a European common arena for discussing Higher Education success issues. [ABSTRACT FROM AUTHOR]

This paper combines a theoretical-based model with a data-driven approach to develop an Early Warning System that detects students who are more likely to dropout. The model uses innovative multilevel statistical and machine learning methods. The paper demonstrates the validity of the approach by applying it to administrative data from a leading Italian university. [ABSTRACT FROM AUTHOR]