A Machine Learning Approach for Predicting Student Enrollment in the University

The Matura exam is the final national examination that high school students in many countries must pass to be eligible for admission to a university. This paper discusses the key factors that have the most impact in producing a reliable machine learning model for predicting students' enrollment in the university. These factors include the final grades from each high school year, matura exam results and the university entry exam points. It should be noted that demographic factors were not taken into consideration in this study. Four machine learning (ML) techniques with a total of sixteen algorithms were implemented using the Weka software: Bayes (Bayes Net, Naive Bayes, etc.), Logistic Regression (Logistic and SimpleLogistic), K-Nearest Neighbors (IBK, KStar, and LWL) and Decision Tree (J48, Random Forest, RepTree, etc.). According to the results, The RepTree algorithm performed the best with a True Positive (TP) rate of 0.902 and a False Positive (FP) rate of 0.148. The algorithm with the lowest performance was NaiveMulti with a TP rate of only 0.790. However, the range between the worst and the best-performing algorithms was 14.18%.