STUDENTS WITH DISABILITIES AND HIGHER EDUCATION ATTAINMENT: PREDICTORS AND ARTIFICIAL INTELLIGENCE

Abstract

Support actions appropriate to the needs of students with disabilities in Higher Education to complete the course are fundamental. Prediction studies, through artificial intelligence, in the educational context for this audience can contribute to improving inclusion policies and minimize academic, social, economic, and political damage caused to all involved. It sought to identify the predictive factors for the completion of Higher Education of students with disabilities through artificial intelligence. This is an exploratory, retrospective study, with a database composed of 563 students with disabilities enrolled in undergraduate courses from 2001 to 2020. Input variables: sociodemographic and academic variables indicated at the time of enrollment. The accuracy of five algorithm models was tested to identify the one with the best performance. XGBoost was the model with the best performance in identifying the predictor variables for higher education completion (ACC=76.38%). The SHAP method of post hoc interpretation was used to identify the degree of importance, the characteristics of each of them, and their relationship with the outcome variable. The predictor factors for completing the course for this audience were high school modality, form of admission, grade in the selection process, age, and gender. The prior identification of predictors for non-completion of the course by students with disabilities can be an important tool that helps the institution to direct actions and resources to meet the needs of these students early and contribute to their permanence, completion of the course, and future occupational perspective.