Early Prediction of Electronics Engineering Licensure Examination Performance using Random Forest

Abstract

Graduate's success on licensure examinations has a significant impact on various facets of a higher educational institution. Using a comprehensive data mining process, this research compared the accuracy of multiple classification algorithms to determine predictors of students' professional certification performance. The Random Forest model achieved the best cross-validated accuracy score of 92.70% based on the evaluation data. A model inspection method of permutation feature importance was used to uncover information from 500 graduates of Southern Luzon State University's electronics engineering program from 2014 to 2019. Among the 33 variables examined, the verbal reasoning or reading comprehension ability of students unveils a clear attribution with their licensure test results along with ratings from different courses in mathematics, professional, and electrical circuits. Thus, the data-driven information can be used to develop programs, initiatives, and techniques to improve success on the electronics engineering licensure examinations.