Abstract: The researches on Data Envelopment Analysis are usually stepping up to follow the cognition level to variable characters. In the early research on DEA,all input and output variables are regarded as quantitative,discretionary and single-role. Now researches have recognized that some variables are non-quantitative,non-discretionary or dual-role. Then different DEA models are advanced to deal with these special variables. This paper reviews the research on the variable classification in DEA contexts.

Key words: Data Envelopment Analysis(DEA), efficiency evaluation, variable classification