Research On The Improvement And Application Of Multi-objective Evolutionary Algorithm

There are a lot of multi-objective optimization problems in the field of science and engineering,and many conflicting objectives need to be optimized simultaneously.For the optimization of the overall goal,we need to consider the interaction and mutual restraint between the sub-goals.The traditional single objective optimization method is not suitable for dealing with multi-objective problems.It is a multi-objective evolutionary algorithm which is now used to deal with multi-objective problems.The so-called evolutionary algorithm is a bionic algorithm that simulates natural selection and evolution of organisms.This kind of algorithm makes full use of fitness function and constraint function without any other prior information,and has strong generality.This paper mainly researched on the application and algorithm improvement of NSGA-II in 0/1 knapsack problem,vehicle routing problem and feature selection.The knapsack problem must satisfy the constraints of capacity.However,because each iteration of the evolutionary algorithm is a random behavior,each iteration produces a large number of infeasible solutions.In this case,in order to make better use of the generated solution,and to get better distribution,this paper proposes a weighted restoration strategy based on the extent of individual violation.At the same time,the strategy is embedded in NSGA-II.Experiments show that the strategy has obvious advantages compared with other repair strategies.Vehicle routing problem refers to a series of distribution centers(or receiving points),scheduling all vehicles,forming a number of driving paths,so that the vehicle can deliver the goods to the specified customer position at the specified time.The vehicle routing problem is also a complex problem containing multiple constraints.The basic framework of this algorithm is NSGA-II.Aiming at the characteristics of this problem,the key operations of the chromosome coding strategy,cross and mutation are improved to adapt to the population evolution,and the overall efficiency of the algorithm is improved.Finally,in order to verify the effectiveness of the algorithm,compared with the ant colony algorithm on multiple test problems,the experiment shows that the algorithm is superior to the ant colony algorithm on multiple test problems.For face recognition problem,this paper selects some main features based on ASM model to do recognition.In order to find the main features more fully,this paper presents the optimization problem model of two target optimization and three targets optimization,and uses the evolutionary algorithm to solve these problems.Through the comparison of several experiments,it is found that the comprehensive performance of the three objectives evolution algorithm is the best.