Student dropout is a challenge for any educational institution, affecting both students and the institution itself. At the University of Oviedo, aware of the scale of the issue, they decided to address it in an innovative way.
The reasons behind dropout can vary widely: academic difficulties, financial problems, personal circumstances, among others. This issue not only has a negative impact on students’ lives, but can also affect the reputation and effectiveness of educational institutions.
A critical challenge and an innovative solution
The University of Oviedo faces a dropout rate it aims to reduce. Although its figures are similar to those of the rest of the Spanish University System, the institution is seeking effective solutions to predict and prevent these situations. This is how, within the framework of the Microsoft Open Education Analytics programme, a unique opportunity arose to explore the capabilities of the Azure platform and its components, leading to a collaboration with ENCAMINA.
The solution proposed by our team consisted of developing a machine learning model capable of analysing students’ historical data to predict the likelihood of dropout. This model makes it possible to identify at-risk students early enough to implement preventive measures. As explained by Alberto Manuel Fernández, Director of the Information Technology Area at the University of Oviedo:
«We found the proposal interesting and, after considering several ideas, developing a dropout predictor based on our historical academic records seemed both attractive and feasible.»
Cutting-edge technology at the service of education
The project was implemented entirely on Azure. Using Azure Synapse for data ingestion, Databricks for processing, and Azure Machine Learning for algorithm development, a robust and efficient model was created. The results are visualised through Power BI, enabling academic leaders to access reports and make informed decisions.
«From Azure, daily data increments are consumed and processed with Azure Synapse, deposited in a Data Lake, and then processed with Databricks by running the Machine Learning algorithm that generates the prediction. This is later visualised with Power BI,» says Alberto Manuel Fernández.
From pilot to real-world application
The model’s accuracy is evaluated on a daily basis. The selected algorithm is retrained every day using 15 years of historical data. This approach makes it possible to continuously adjust the model and generate reliability indicators. According to Alberto Manuel Fernández:
«The model is quite accurate in determining who is NOT going to drop out, and somewhat exaggerated when predicting who WILL drop out. On average, it has a 0.95% error rate in non-dropout predictions and overestimates dropout predictions by 60%.»
Although the tool is still in the pilot phase, it is expected to complement the tutorial action plans of all faculties and schools, helping to reduce university dropout through earlier and more precise interventions. Early detection is crucial, especially during the first terms of the first academic year, although this also presents challenges due to the limited activity records available for new students at that early stage.
According to the IT Director at the University of Oviedo
ENCAMINA has been a strategic partner throughout the entire process, from presenting the initiative to developing the project. Its support has been key to the success of this initiative, providing not only the technical infrastructure but also the guidance needed to implement and fine-tune the model.
The University of Oviedo, with the support of ENCAMINA and Microsoft, has taken a significant step towards reducing university dropout through advanced technology. This project stands out not only for its innovation, but also for its potential to transform educational management and significantly improve students’ lives.
¡Suscríbete a nuestras RRSS!
