We study the new concepts of grade evaluation which is dominated by students themselves instead of by the instructors. Each individual student, at the end of the class, determines her/his weight ratios for all of her/his subgrades in order to optimize her/his final grade. However, by considering the fairness, each individual-determined ratios should have some degree of compromise with other students'. On the other hand, we should no longer strive for a normal distribution for the grades. In this study, based on the concepts of data envelopment analysis (DEA), we developed a linear programming model to our problem. The reason is that the inherent property of compromise in DEA matches with ours. We apply our linear programming model to the real example and the results shown are highly consistent with what we expected.