Early identification of academic risk in engineering students using machine learning

Abstract

This study describes the development of a predictive model designed to identify students at pre-risk and academic risk stages within the Faculty of Engineering of the Universidad de Atacama. The primary objective is to anticipate factors that may contribute to dropout or low academic performance, thereby enabling the implementation of timely and targeted support interventions. A sample of 130 first-year students from the Civil Engineering in Computer Science undergraduate program was used. The methodology involved collecting and integrating academic data from institutional systems, including variables such as failed courses, withdrawn courses, academic level, and curricular progress. Machine learning techniques were applied to identify patterns that could predict academic risk. The model’s results enabled the early detection of students in pre-risk situations, facilitating targeted interventions such as reduced academic workload, personalized tutoring, and active monitoring. These actions contributed to improved academic performance, increased student engagement, and reduced dropout rates. This study demonstrates that integrating predictive analytics technologies with instructional support strategies can enhance retention and academic success from the early stages of university education.