A comparison of artificial neural networks and the statistical methods in predicting MBA student’s academic performance

Abstract

MBA has become one of the most popular and vital professional degrees internationally. The MBA program admission process’s essential task is to choose the best analysis tools to accurately predict applicants’ academic performance potential based on the evaluation criteria in making admission decisions. Prior research finds that the Graduate Management Admission Test (GMAT) and undergraduate grade point average (UGPA) are common predictors of MBA academic performance indicated by graduate grade point average (GGPA). Using a sample of 250 MBA students enrolled in a state university with AACSB accreditation from Fall 2010 to Fall 2017, we test and compare the effectiveness of artificial neural networks (ANNs) against traditional statistical methods of ordinary least squares (OLS) and logistic regression in MBA academic performance prediction. We find that ANNs generate similar predictive power as OLS regression in predicting the numerical value of GGPA. By dichotomizing GGPA into categorical variables of “successful” and “marginal,” we identify that ANNs offer the most reliable prediction based on total GMAT score and UGPA while logistic regression delivers superior performance based on other combinations of the predictors. Our findings shed light on adopting ANNs to predict academic performance potential with a strong implication in MBA admissions to select qualified applicants in a competitive environment.