Strategy Of Interval Multi-objective Optimization Integrating Cloud Model And Information

In practical applications,optimization problems often contain uncertainty of objective and subjective and interval multi-objective optimization problem is a very important class of uncertain optimization problems.This problem not only contains multiple objectives,and the target function value is the interval number.What's more,the precise function of the interval multi-objective optimization is extremely difficult to obtain.All of these factors increase the difficulty of resolving the optimization problem.Such issue is summed up as “the problem of interval multi-objective optimization with unknown optimization functions”,in this paper,the idea of cloud model and data mining is introduced into NSGA-II algorithm to solve this kind of problem.For the problem that the optimization function is known,a kind of non-dominated sorting cloud model algorithm for solving interval multi-objective optimization is proposed,in which cloud model is used to improve the NSGA-II algorithm.Firstly,genetic operator such as crossover and mutation in conventional NSGA-II is replaced by normal cloud operator,and offspring cloud droplets derived from random parent in initial cloud cluster.Secondly,generated cloud droplets are disposed and kept based on constraint conditions,so infeasible solution doesn't be left in the next algorithm.Finally,the feasible solution is sorted by interval dominant relationship,and then the crowding distance of incomparable same order solution is calculated.For the problem that the optimization function is unknown,a NSGA-II algorithm is proposed to solve the problem of the right dominates relations of population individual for the interval multi-objective optimization with unknown optimization functions,which integrated Gaussian surrogate model and nearest neighbor.Firstly,Gaussian surrogate model is built up by training sample sets and the super-parameters are calculated by genetic algorithm,therefore,the possibility degree probability between candidate solutions are obtained according to Gaussian surrogate model;secondly,the similarity between candidate solutions and sample solutions is calculated by nearest neighbor,thus the possibility degree probability is obtained;finally,the regression memory method is dynamically used to adjust the weight of dominant results between Gaussian surrogate model and nearest neighbor to work the dominance relations of population individual out adn the improved NSGA-II algorithm work the problems out.