Research On Multi-objective Optimization Problem Based On Evolutionary Computation

In the real world,various optimization problems exist on engineering projects,scientific research and daily life.Multi-objective optimization problem(MOOP)is widely applied to various fields,such as network structure design,workshop flow scheduling,production scheduling,etc.The travel salesman problem and the community detection problem are one of the representative examples of the application of multi-objective optimization problems with graph theory and complex network fields,respectively.(1)In order to solve the multi-objective traveling salesman problem based on genetic algorithm,GAbased algorithms suffer the premature convergence,the insufficient diversity,and nonuniform distribution of solutions.In order to overcome these problems,this paper proposes an improved method of genetic algorithms based on a novel evolutionary computational model,named as Physarum-inspired computational model(PCM).This method can be used to solve a variety of practical optimization problems which can be transformed into the travel salesman problem with multiple optimization objectives.First,based on the prior knowledge that PCM can find the shortest path of two source points,the population initialization of the proposed method is optimized,so as to enhance the ratio of potential solutions and the quality of the initial population.The optimization strategy can effectively improve the convergence speed and the distribution of the solution to achieve the optimal solution.Then,in order to balance the convergence and diversity of pareto solutions,the hill climbing method is added to the genetic operation as a local operator,in order to improve individual diversity and expand exploration space to avoid falling into local optimization.Simulation experiments are carried out on different artificial datasets and real-world datasets,respectively.The experimental results show that the proposed algorithm in this paper can obtain a better pareto front with the wider distribution range and better solution quality,compared with other state-of-the-art algorithms.The proposed algorithm shows better convergence speed,local search ability and diversity retention,so it is easier to find the global optimal solution.(2)In order to effectively balance the community structure of each layer to obtain a high-quality composite community in multilayer network,this paper transforms the multi-layer network community detection into a multi-objective optimization problem and put forward a new GA-based multi-objective optimization algorithm NSGAMOF for multi-layer network community detection.In order to overcome the fact that some existing optimization-based algorithms are easy to fall into the local optimality and are difficult to be applied to high-dimensional networks,the objective function of each layer in the multi-layer network is optimized as one objective in NSGAMOF.In addition,the NSGAMOF iteratively optimizes each objective function.The local search is designed as a local operator embedded in genetic operations in order to overcome the local optimal solution.Especially,by evaluating the objective function to form the destination field,three different optimal solution selection strategies determine the optimal composite community.Experiments on both real-world and synthetic networks demonstrate that the NSGAMOF algorithm shows better performance and higher than other state-of-the-art algorithms in multi-layer networks.Particularly,by changing layer number and the network structure,the algorithm can still detect high-quality communities,which effectively avoids falling into a local optimal solution.Experiment results show that the NSGAMOF algorithm can be used in higher dimensional multi-layer networks.