Evolutionary Algorithms For Multi-objective Optimization Problems Based On Classification Design

In various fields of daily life and scientific research, there are many kinds of optimization problems. Solving these optimization problems means looking for the most reasonable and most reliable solutions from all possible solutions. For the multi-objective optimization problems, objective functions are usually conflict with each other and their optimal solutions (Pareto solution) are often not unique, instead, there are usually many Pareto optimal solutions or infinite number of Pareto optimal solutions. How to find a number of representative and uniformly distributed Pareto optimal solutions is a difficult task. To design efficient and effective algorithm, it's very important to make full use of the information of every objective function. In the existing optimization algorithms, EA has become an acknowledged method to solve the optimization problems regarding its potential to solve multi-objective optimization problems.In this paper, two new optimization algorithms based on the classification design and genetic algorithm will be proposed. Numerical simulations demonstrate the efficiency and effectiveness of the proposed algorithm. This paper mainly includes:First, the classification function and its standard based on objective functions are defined. A simple kind of classification function is designed, and a new algorithm called CL-MOEA based on the uniform design is presented. This algorithm mainly consists of classification function design and the evolutionary operators design.Second, the initial population is generated by reasonably combining clustering method with uniform design, and a simple evolutionary operator is designed. Based on these, a new clustering-evolutionary algorithm called C-MOEA is proposed. The algorithm highlights advantages of the clustering method in dealing with population classification.Finally the numerical experiments demonstrate the effectiveness of the twoproposed algorithms.