Student dropout rates can be a significant worry when it comes to ensuring success in higher education. In a recent study co-authored by our researchers Pilar Aparicio-Chueca and Xavier M. Triadó-Ivern published in Studies in Higher Education, delved into the world of university life to uncover the key factors predicting student dropout.

The study focused on 3,583 first-year students pursuing a Business Administration (BA) degree at the University of Barcelona. The findings highlighted two crucial variables that played a pivotal role in predicting student dropout: the percentage of subjects failed and the number of classes not attended during the first semester.

To confirm the robustness of these results, the study extended its scope to include an additional 10,784 students from three different degree programs—Law, BA, and Economics—at the Complutense University of Madrid. The researchers employed three different algorithms—neural networks, random forest, and logit—to analyze the data.

In the case of neural networks, the study utilized the NeuralSens methodology, which is based on the use of sensitivities, allowing for a more interpretable analysis. The outcomes consistently showed that both simpler models (like logit) and more sophisticated ones (neural networks and random forest) exhibited high accuracy in predicting student success and dropout. On average, these models achieved accuracy rates of 77% and 69%, respectively, in the test sets.

One noteworthy aspect of the study was its reliance solely on academic data from the universities themselves to develop the models. This approach ensures that the predictions are not influenced by other personal or organizational variables, which are often challenging to access.

In essence, this research provides valuable insights into the factors that can make or break a student’s journey through higher education. By understanding and addressing these variables, educators and institutions can better support students, increasing the likelihood of success and reducing dropout rates.