Employability Prediction of Engineering Graduates Using Ensemble Classification Modeling

Higher educational institutions have a responsibility and commitment to deliver employable graduates as it impacts their well-being and the economy. This study compared the accuracy of several classification algorithms to build an ensemble prediction model capable of forecasting graduates' employability using extensive data mining techniques. Based on the evaluation metrics, an ensemble model composed of Random Forest (RF), Support Vector Machines (SVM), and Naïve Bayes (NB) achieved the highest cross-validated accuracy score of 93.33%. Association rule mining and permutation feature importance analysis from 500 graduates of the electronics engineering program of a university revealed that grit is firmly attributed to employability, including the capabilities to acquire technical skills and professional certifications. Thus, the knowledge gained can be used to develop a range of policies, initiatives, and strategies to increase students' employment prospects.