Using academic analytics to predict dropout risk in engineering courses
Conference Paper
Overview
Research
Additional Document Info
View All
Overview
abstract
The increase of data generated and stored in the educational databases makes it possible to obtain essential
information about the teaching and learning process. School dropout and performance problems continue to represent
issues which challenge teachers, researchers and higher education institutions to seek solutions. Through the use of
academic analytics techniques for data analysis, a sample of 1,844 students between graduates and dropouts on the period
between 2007 and 2015 were used as the basis. The methodology followed is essentially quantitative and it allowed to
compare student profiles and degrees based on scores, number of attempts and other performance indicators. The data set
was processed using Excel software for statistical analysis and R software for data mining using the k-Means and C5.0
algorithms. The propose of a model based on decision trees has as main objectives the generation of standardized
instructions, easy interpretation and allow the addition of several possible outcomes, contributing to the decision-making
process. The results of this study resulted in contributions which enable higher education institutions to identify students
with performance problems and those at risk of dropout and, therefore, allow teachers and course directors to adopt better
strategies to increase success and reduce dropout.
The increase of data generated and stored in the educational databases makes it possible to obtain essential information about the teaching and learning process. School dropout and performance problems continue to represent issues which challenge teachers, researchers and higher education institutions to seek solutions. Through the use of academic analytics techniques for data analysis, a sample of 1,844 students between graduates and dropouts on the period between 2007 and 2015 were used as the basis. The methodology followed is essentially quantitative and it allowed to compare student profiles and degrees based on scores, number of attempts and other performance indicators. The data set was processed using Excel software for statistical analysis and R software for data mining using the k-Means and C5.0 algorithms. The propose of a model based on decision trees has as main objectives the generation of standardized instructions, easy interpretation and allow the addition of several possible outcomes, contributing to the decision-making process. The results of this study resulted in contributions which enable higher education institutions to identify students with performance problems and those at risk of dropout and, therefore, allow teachers and course directors to adopt better strategies to increase success and reduce dropout.
