Research And Application Of Multi-objective Evolutionary Algorithms With Uncertain Preference

Multi-objective optimization problems have always been a hot issue in the field of science and engineering research.The optimal solution of the multi-objective optimization problem is not a single optimal solution,but a set of optimal solutions consisting of Pareto optimal solutions.With the increase of the objective number,the search mechanism of traditional multi-objective evolutionary algorithms quickly falls into the bottleneck,and the performance of the algorithm declines drastically,which makes it face a great challenge:the number of solutions required to cover the entire Pareto front grows exponentially.In addition,the proportion of non-dominated solutions in the population increases dramatically,and the difficulty of visualization of the solutions is greatly increased.In order to solve the above difficulties,scholars consider using decision makers'preference information to deal with multi-objective optimization problems.By combining the preference information of decision makers with the multi-objective evolutionary algorithm,the search of the algorithm can be concentrated in the preference region that the decision makers are more interested in.This not only can use the computational resources of the algorithm effectively,improve the efficiency of the algorithm,but also can reduce the computational complexity.However,in the actual optimization problem,decision makers often cannot give clear preference information because of the lack of prior knowledge.At the same time,the performance of the traditional preference multi-objective optimization algorithm is seriously affected by the position of the reference point.When the reference point is located at some extreme positions,it may cause degradation or even non-convergence of the algorithm.In order to solve the above problems,this paper introduces uncertainty preference information into multi-objective evolutionary algorithms to effectively solve the many difficulties faced above.In this paper,implicit preference information is used as the starting point,and the angle preference strategy is proposed.The knee point is used as the implicit preference point,and the angle value is used to control the range of the preference region to guide the population to search the preference region.At the same time,the replacement optimization strategy is proposed.By finding and replacing the difference solution in the current neighborhood,the problem of insufficient convergence of the algorithm in the replacement process is alleviated.Furthermore,a similar mapping method based on implicit preference is proposed.According to the historical knee point,the K-means clustering method is used to obtain the central point,the implicit preference vector is determined,and the similarity mapping centered on the implicit preference vector is constructed.The search of the algorithm is concentrated near the implicit preference vector,and the influence of the position of the reference point on the performance of the algorithm is eliminated.At the same time,the neighbor matching strategy is proposed.The sub-problem of the individual is assigned to the nearest individual,and the uniformity of the algorithm in the iterative process is maintained,thereby improving the overall performance of the algorithm.Finally,the algorithm proposed in this paper is applied to solve the resource allocation problem of software testing,which provides a better choice for decision makers,and provides new ideas and new methods for solving multi-constrained nonlinear multi-objective optimization problems with practical engineering background.