In data envelopment analysis (DEA) modelling applications, analysts typically experience difficulty in choosing variables when the number of variables is greater than the number of decision-making units (DMUs). In this paper, we develop a novel method to facilitate variable selection in DEA using entropy theory to avoid information redundancy. A numerical analysis is provided to compare our method to those of related studies. The results show that our proposed method produces a lower Akaike information criteria (AIC) value than other approaches. By presenting a real-world case, we show that this new method yields useful managerial results.