Abstract: The multiple attribute comprehensive evaluation models which could deal with three types of index(benefit, cost, fixed) were constructed based on data envelopment analysis(DEA). The cross efficiency concept was introduced to link the self-evaluation with peer-evaluation of decision making unit(DMU) for solving the problem that optimal weights of the DEA based multiple attribute comprehensive evaluation model are non-uniform. A new game cross efficiency concept was combined with the DEA based model to deal with that optimal weights of the DEA based model are non-unique. The game cross efficiency concept, in which each evaluated DMU is seeking to maximize its own efficiency under the condition that the cross efficiency of the other DMU does not deteriorate, enhanced the discrimination and rationality of the evaluation results. The DEA based model was validated by a numerical example and was applied to the project review of equipment design which indicates that the method is efficient in solving the project evaluation, rank and selection problems.