Predictive models of student desertion at a private Peruvian university

Abstract

Desertion is a problem that affects public and private universities, and leads to a series of negative consequences for both institutions and students. Therefore, the objective of this study was to determine how the use of predictive models in low pass-rate courses helps to identify students at risk of desertion. Seven predictive models were designed using CRISP (Cross- Industry Standard Process for Data Mining) methodology and students’ academic records to be applied in seven low pass-rate courses. Among the main results, it can be noted that predictive models contributed to the reduction of fail rates by 25% and 40%, and that the variables that best forecast desertion were career choice (vocation), number of times students enrolled in the course, and grades obtained in mathematics or language arts when students attended the fifth year of high school.