Cooperative coevolution has earlier proven to be an effective means to target classification. It is the aim of this paper to further investigate if a supplementary archiving mechanism, which would preserve a variety of the best evolved rules, would boost the accuracy of prediction of the proposed technique. The output of the coevolutionary algorithm would consequently consist of several different prototypes for each outcome of the classification task which would subsequently lead to an enhancement of the performance and generalization ability of the classifier. The novel approach is tested against two real-world decision problems regarding tumor diagnosis and obtained results sustain the employment of the additional archive.