Improving Performance of Rule-Based Predictive Models Using Ensemble Learning Technique for Predicting Graduates' Employability

Abstract

This research proposes to compare the efficiency rule-based predictive models using ensemble learning techniques to improve the predictive performance for graduates' Employability. Four rule-based techniques were used: MODELM, PART, J48, and Random Trees, and predictive performance was improved using ensemble learning techniques based on Boosting Method (AdaBoost). Moreover, 10-fold cross-validation was utilized to split the data into the training and test set. This research has measured performance models with accuracy, precision, recall, and f-measure. The experimental result demonstrated that improving the performance of rule-based predictive models using ensemble learning techniques can increase performance for all techniques, that the accuracy increased by an average of 6.99%, the precision increased by an average of 6.90%, the recall increased by an average of 7.05%, and the f-measure increased by an average of 7.15%. The predictive model from the J48 algorithm combined with AdaBoost highest efficiency with accuracy 89.47%, precision 89.50%, recall 89.50%, and f-measure 89.50%.