For Certus, one of Peru's largest higher education institutes, I was tasked with developing a predictive model to address the critical issue of student dropout, which had become increasingly pressing in the wake of the pandemic. The educational sector was significantly impacted, with both new enrollments and retention of current students seeing a sharp decline. One of the institution's primary concerns during this period was to maximize the retention rate, particularly among new students.

In response to this challenge, at the request of the institute's authorities, I embarked on creating a model capable of identifying incoming students who were at risk of dropping out. This project was significant not only for its immediate implications for student retention strategies but also for its potential to shape the institute's approach to educational support and intervention in the long term.

The development of this predictive model was facilitated by the collaboration with the institution's IT department, which provided access to a comprehensive dataset encompassing demographic, academic, and vocational information about the incoming students. This rich dataset was crucial for understanding the multifaceted nature of student dropout risk and for designing a predictive model tailored to the institute's specific context.

Employing machine learning (ML) techniques, specifically classification models, we aimed to create a robust and reliable tool for predicting dropout risk. The use of ML-based classification models was a strategic choice, leveraging the potential of these technologies to uncover patterns and insights within complex datasets that traditional analytical methods might miss.

The resulting model achieved a level of sensitivity significantly higher than that of a dummy model, which would make predictions randomly. This achievement was a testament to the model's effectiveness and its suitability for the institution's needs. The superiority of the predictive model over a random approach indicated its potential to provide actionable insights for the institute, enabling targeted interventions to support at-risk students.

This project was not just about leveraging technology to solve a problem; it was about addressing a critical social issue within the educational sector. The successful development and implementation of the predictive model demonstrated the potential of data-driven strategies to make a meaningful difference in students' lives, especially during challenging times. The model's success laid the groundwork for future initiatives aimed at enhancing student retention and success, marking a significant step forward in the institution's efforts to support its students through data-informed decision-making.

Technical stack:

Data Sources:

Processing:

Visualization: