In Colombia, the desertion average rate shows that only around half of the students that begin an undergraduate program are finishing their studies. Although several models have been developed, the results of the implementation of strategies for student retention have not been sufficiently effective. Therefore, this study focuses on the creation of robust predictive models that allow timely anticipation of the risk that an engineering student will retire prematurely from the program. This phenomenon is analyzed in an empirical way through a methodology of Knowledge Discovery in Databases (KDD) using different machine learning techniques. Results show that the academic cost (in terms of the number of subjects viewed) and the semester the student entered have a significant impact on the probability of dropout occurring, especially when considering the dropout in the firsts semester. Then, acting on the students predicted by the model might reduce the number of dropouts.
|Title of host publication||Proceedings of the 8th Research in Engineering Education Symposium, REES 2019 - Making Connections|
|Number of pages||8|
|Publication status||Published - 2019|
|Name||Proceedings of the 8th Research in Engineering Education Symposium, REES 2019 - Making Connections|
Bibliographical noteFunding Information:
The authors wish to acknowledge the financial assistance of Universidad Sergio Arboleda. The authors also acknowledge the support of Juan aD vid Arboleda and María Paula Flórez during the data gathering and research.
Copyright © 2019 Andrés Acero, Juan Camilo Achury and Juan Carlos Morales.