Student desertion, factors associated with decision trees: the case of a graduate school at a public university in Peru

Díaz B.; Marín W.; Lioo F.; Baldeos L.; Villanueva D.; Ausejo J. · RISTI - Revista Iberica de Sistemas e Tecnologias de Informacao (2022)

Resumen

The objective of the research was to predict the behavior of student desertion using the technique of data mining with decision trees, through the J48 algorithm developed in the free software WEKA (Waikato Environment for Knowledge Analysis) solution that was applied to the master’s degree programs in education of the Graduate School at the Universidad Nacional José Faustino Sánchez Carrión, Peru; it was possible to identify associated characteristics that influence the main factors such as socioeconomic and academic factors. The research was technological, applied, of temporal scope; a 20-item questionnaire was applied to 237 enrolled students for data collection. The result obtained was the prediction of student desertion using the J48 WEKA supervised automatic learning algorithm by means of decision trees, thus identifying the main factors that influenced desertion, with the J48 algorithm an accuracy of 87. 76% and 66.63% of moderate concordance through Cohen’s Kappa index, concluding that from the results obtained, the managers identify the students with possible risk of dropping out of their studies and thus take corrective measures and apply strategies to help reduce the student dropout rate.

Instituciones: Universidad Nacional José Faustino Sánchez Carrión