PREDICTIVE Modeling of Student Retention Across Different Student Cohorts Using Machine LEARNING

Abstract

This study aimed to predict student retention using machine learning, particularly logistic regression, integrated into the Guidance Office Records Management System (GORMS) of the participating catholic institution in Zamboanga City. It focused on system development, predictive performance, technology acceptability evaluation, and integration planning. A quantitative research design with a hybrid model as a system development approach was utilized. Student records were used to develop the logistic regression model, while system users, including guidance personnel, administrators, and IT experts, evaluated the system’s technology acceptance using an adopted instrument. Data were analyzed using mean and standard deviation. The application was fully developed in PHP and Python. Major modules were treated as prototypes, including student profiling, student retention prediction, the institutional integration survey, and the exit interview. User-specific access rights were implemented to ensure data privacy. Logistic regression showed satisfactory predictive performance, as indicated by accuracy, recall, precision, and F1-measure. A train-test split and k-fold cross-validation were used to evaluate the model's performance. Furthermore, the technology acceptance evaluation revealed very high levels of acceptance across perceived usefulness, perceived ease of use, and intention to use, suggesting a strong likelihood of adoption. An integration plan was proposed, with the institution serving as the pilot school and future implementation across RVM schools in Southern Mindanao. The institution and other RVM schools will be able to use the results to implement early interventions to improve student retention.