2025 ASEE Annual Conference & Exposition

This research article proposes developing a predictive model to identify, at an early stage, students at risk of low academic performance. Academic performance is a critical indicator in higher education, essential for student and professional success, and has been extensively studied due to its impact on retention and timely graduation. Socio-demographic factors such as socioeconomic status, family environment, work responsibilities, study habits, financial support, and psychological factors have been shown to influence student performance and attrition rates significantly. While progress has been made using linear regression models, recent years have seen the incorporation of advanced artificial intelligence techniques, offering new opportunities to enhance academic management. The objective of this article is to design a predictive model based on the entry profile of engineering students to assess their risk of low academic performance. The study employs a non-experimental quantitative methodology and machine learning techniques within a Knowledge Discovery in Databases (KDD) framework. The data used in the model includes Weighted Average Grades and socio-demographic factors from the characterization survey that students complete upon entering the university. The sample comprises 1,266 students from the Faculty of Engineering at a private university in Chile who enrolled in the first semester of 2022. Their academic performance is analyzed from that semester until the first semester of 2024, covering five semesters and reaching 50% of their curricular progression. Additionally, socio-demographic data such as family, economic, and work backgrounds, collected in the characterization survey conducted at the time of their entry in 2022, are utilized. The results of this research are expected to identify key factors affecting academic performance, such as the number of working hours, study methodologies, and the source of financing for their studies. The developed model is anticipated to classify academic outcomes into four performance levels: no mastery, insufficient mastery, satisfactory mastery, and outstanding mastery, with 98% accuracy and a 3% margin of error. This predictive model aims to contribute to academic management by facilitating the early implementation of support measures or programs for students at academic risk. Moreover, institutions can adjust curricular design and teaching methods by analyzing the factors influencing academic performance. The actions derived from this model are expected to improve students' academic performance, potentially reducing dropout rates and increasing timely graduation rates, thereby inspiring and motivating educators and policymakers in engineering education.