ABSTRACT

This paper proposes a TSK-type neural fuzzy network (TNFN) with a group interaction-based evolutionary algorithm (GIEA) for constructing the cancer cell colonies diagnosis system (CCCDS). The proposed GIEA is designed on the basis of symbiotic evolution which each chromosome in the population represents only partial solution. The whole solution consists of several chromosomes. The GIEA is different from the traditional symbiotic evolution. Each population in the GIEA is divided into several groups. Each group represents a set of the chromosomes that belongs to only one fuzzy rule. Moreover, in the GIEA, the interaction ability is considered that the chromosomes will interact with other groups to generate the better chromosomes by elites-base interaction crossover strategy (EICS). In the CCCDS, the EICS is used to train the CCCDS. After trained by the EICS, the CCCDS can diagnose the visible cancer cell colonies automatically. The performance of the GIEA is proved to be better than other existing models in diagnosing cancer cell colonies.