Explaining and Predicting Graduate Employability Using Data-mining Techniques

Abstract

Data-mining has been applied in various areas because of its ability to rapidly analyze vast amounts of data. For a higher education institution with a huge data of graduate employability, data-mining technique is an appropriate for analyzing this information. This study presents a graduate employability model that uses a classification task to identify the most important causes of graduate employability, and to compare the accuracy between Bayesian methods (Naïve Bayesian, Averaged One-Dependence Estimators, Averaged One-Dependence Estimators with subsumption resolution, Naïve Bayesian Simple, Bayesian networks, and Naïve Bayesian Updateable) and Decision tree methods (J48, J48 Graft, Simple Cart, Random Forest, Logical Analysis of Data tree, Rep tree, Decision Stump, and Random tree). The experiments show that 3 attributes with a direct effect on employability are the work province, occupation type, and times find work. The results can provide valuable insights for domain experts. Moreover, the results of this study show that the J48 Graft algorithm, which is a variant of the Decision tree method, had a higher accuracy than the average of other Bayesian algorithms.
Copyright © 2014 Praise Worthy Prize - All rights reserved.