Based-decomposition Multi-objective Genetic Algorithm And Its Application In Community Detection

Multi-objective optimization algorithm is widely used in real life,and it's favored by many researchers because of its good computing performance.MOEA/D are the focus of research at present.But algorithm lacks the selection pressure of elite solution for optimization problems with complex Pareto front(PF).In addition,with the increase of the number of objective function and variables,the number of optimal solution sets is also increasing.It makes the number of solutions approaching the PF increase exponentially,which weakens the ability of the algorithm to search for the global optimal solution.Based on this,a new algorithm is designed by using weight vector and neighborhood adaptive adjustment mechanism.On this basis,a dimension reduction genetic algorithm based on domain knowledge guidance is proposed.Finally,the proposed algorithm is used to solve the community detection problem of complex network.Aiming at the deficiency of MOEA/D,this paper designs weight vector and neighborhood adaptive adjustment mechanism.Firstly,a uniform vector generation method is proposed,which improves the uniformity of solution set of multi-objective algorithm.In addition,a neighborhood adaptive adjustment mechanism is proposed,which changes from the global search evolution to the local search evolution gradually.Then,according to the position of the weight vector in the hyperplane,the neighborhood is secondary adjusted.The algorithm is applied to several benchmark optimization problems,and experiments show that algorithm performs well compared with the current popular genetic algorithms.For large-scale many-objective optimization problems,a dimension reduction genetic algorithm based on domain knowledge guidance is proposed in this paper.Firstly,a dimensionality reduction method of objective function based on correlation is proposed to cluster the objective functions with high correlation.On this basis,a domain knowledge-guided genetic algorithm is designed to solve the optimization problem after dimensionality reduction.To obtain a high-quality solution set,algorithm also presents a population initialization method based on the mirror-partitioning of decision space.At the same time,the algorithm uses the acquired domain knowledge to continuously recruit new excellent individuals.Many comparative experiments are carried out in this paper,and experimental results show that proposed algorithm can solve large-scale many-objective optimization problems(LSMAOPS)more effectively.Finally,proposed algorithm is used to solve the community detection problem of complex networks.This paper proposes an improved genetic algorithm based on domain knowledge in community detection.Algorithm framework adopts the MOEA/D,using simulated annealing algorithm in local search.In addition,combining with the community detection problem characteristics of complex networks,the new function(M)is proposed as the objective function of the optimization algorithm.The LPA-M method is used to initialize the populations.Algorithm improves individual hybrid crossover and mutation strategies.Experimental results show that the proposed algorithm has excellent performance on a series of synthetic and real networks.In this paper,aiming at the shortcomings of MOEA/D,proposed a multi-objective genetic algorithm with the adaptive adjustment mechanism of weight and neighborhood,and a dimension reduction genetic algorithm based on domain knowledge guidance.They greatly improve the efficiency and precision of the algorithm.In addition,the proposed algorithm to solve the community detection problem of complex network.Experimental results show that the proposed algorithm is of important theoretical significance and application.