Modelo predictivo de la mortalidad académica del programa de Ingeniería de Sistemas de la USTA Seccional Tunja basado en técnicas de Machine Learning

Abstract

The project's aim was to develop a predictive model using machine learning techniques, which allowed identifying cases of academic mortality that may lead to student dropout in the computer science program. The work began with a bibliometric analysis to gather data from research that has implemented machine learning in the study of academic dropout. Additionally, inquiries were made about the machine learning techniques that best fit the subject and international experiences. Subsequently, the dataset was consolidated based on the academic information of computer science students from 2018-1 to 2021-2. Afterwards, feature engineering was applied to determine which ones were the most relevant for a dropout predictor. In this way, the machine learning model for predicting students in the computer science program at the university was determined. This process was conducted by comparing performance indices using different machine learning algorithms with the previously collected information. Thereafter, the model was validated using the cross-validation technique. Finally, a dashboard was deployed, showing the factors that influence student academic dropout and the predictions of dropout risk in students.