Setting Predictor Variables to Generate an Early Dropout Prediction Model for First-year University Students. Analysis of a Sample of First-year Students Beneficiaries of the Academic Leveling Scholarship in a Public University in Chile

Authors

  • Cristian Contreras Universidad de Valparaíso

DOI:

https://doi.org/10.31619/caledu.n54.828

Keywords:

Democratization of education, Dropping out, Educational indicators, Educational policy, Diversification of education

Abstract

The quantitative results of the program "Academic Leveling Scholarship" from year 2016 were analyzed, from a sample of 250 beneficiaries. Statistic datasets were checked using Decision Tree and Clustering techniques in the search of information in institutional data bases that will be relevant to develop an Early Dropout Prediction Model. This model will allow for selecting students with a higher risk of dropping out to receive a timely focused accompaniment. It concluded that there exists, at least, one characterization variable upon the first-year students that allows for identifying the dropout risk in the selected sample regarding the score obtained in the specific math test of the university selection process.

The relevance of this research is its contribution to predictive evidence to align efforts that different universities are making to improve their permanence indexes, especially regarding the recent diversification of the population that has been admitted to higher education due to universal access policy, free education, equity and inclusion.

Downloads

Published

2021-07-30

Issue

No. 54 (2021): Revista Calidad en la Educación

Section

RESEARCH PAPERS

License

Authors retain their Copyright and only transfer a part of these to the journal, accepting the following conditions:

Authors keep their rights as authors and guarantee the right to the journal for the first publication of their work, which is simultaneously subject to the Creative Commons Attribution license allowing third parties to share the study accrediting the author and first publication in this journal.

Authors may adopt other non-exclusive license agreements for distribution of the version of the published work (e.g. inclusion in an institutional thematic file or publication in a monographic volume) accrediting initial publication in this journal.

Authors are allowed and recommended to share their work over the Internet (e.g. in institutional telematic files or their website) before and during the submission process, which may lead to interesting exchanges and increased citation of the published work. (See The effect of open access).