Model for Prediction of Student Dropout in a Computer Science Course

This work presents a model that can predict the student’s risk of dropout using data from the first three semesters attended by Computer Science Undergraduate students. Nowadays, Educational Management Systems store a large amount of data from the interaction of not only students and professors but also of students and the educational environment. Analyze and find patterns manually from a huge amount of data is hard, so Educational Data Mining (EDM) is widely used. This work uses the CRISP-DM methodology and data from Computer Science Undergraduate students from Federal University of Pelotas, Brazil. The results are shown for three algorithms: the Decision Tree algorithm presents a precision of 84.80%, a Recall of 85.80% and an AUC of 77.24%; the Random Forest algorithm presents a precision of 88.57%, a Recall of 90.14% and an AUC of 83.22%; the Logistic Regression algorithm presents a precision of 71.24%, a Recall of 94.28% and an AUC of 58.39%. The results indicate that it is possible to use a prediction model using only the data from the first three semesters of the course.