Interactive Genetic Algorithms with Individual's Fuzzy and Stochastic Fitness

We presented a novel interactive geneticalgorithm to effectively characterize a user's cognition onthe evaluated objects and the stochastic phenomena in theevolutionary process. A fuzzy number is adopted to express an individual's fitness to reflect a user's cognition,and a stochastic variable with normal distribution to depict the stochastic behavior in the fitness based on the error theory. The fuzzy number and the stochastic variableare transformed into different intervals with ? cut set leveland confidence level, respectively, and different individuals in the same generation are compared based on intervaldominance. The values of ? and stochastic parameter aredetermined according to the uncertainty degree of a user'scognition which is quantitatively described with the fuzzyentropy of a fuzzy number. Finally, the algorithm wasapplied to a fashion evolutionary design system and theexperimental results show its rationality and effciency.