This study develops a method for identification of risk of desertion of undergraduate students. It provides evidence to predict and reduce factors that influence this risk, and present the result obtained by a Higher Education Institution (HEI). In this study, classified as quantitative, descriptive and survey-type, we applied a structured questionnaire to 8,750 students. As a theoretical contribution, we highlight the identification of dimensions that influence the decision to drop out or stay and the proposed method, which can assist in the development of new researches. Students at risk of desertion were identified using artificial neural networks and cluster analysis. Using this diagnosis, the students in risk of desertion were contacted and followed by the program coordinators. It contributed to 18% reduction in the Institution´s attrition rate. The dimensions with most influence were professional placement and vocation of the student, availability of time for study and personal life factors. Regarding practical contributions, it is noted that, from the presented method, the Institution can identify students at risk of desertion and thus develop strategies and actions for these students to remain in their studies if they wish.
Strategy in higher education institutions; Attrition of students; Propensity; Artificial neural networks; Cluster analysis
