Global Prediction-based Adaptive Mutation Particle Swarm Optimization

Particle swarm optimization (PSO) algorithm has attracted great attention as a stochasticoptimal method due to its simplicity and power strength in optimization fields. However, twoissues are still to be improved, especially, for complex multimodal problems. One is thepremature convergence, and the other is the low efficiency for complex problems.To address these two issues, firstly, a strategy based on the global optimum prediction isproposed. A predicting model is established on the low-dimensional feature space with theprinciple component analysis technique, which has the ability to predict the global optimalposition by the feature reflecting the evolution tendency of the current swarm. Then thepredicted position is used as a guideline exemplar of the evolution process together with and. Secondly, a strategy, called adaptive mutation, is proposed, which canevaluate the crowding level of the aggregating particle swarm by using the distributiontopology of each dimension, and hence, can get the possible location of local optimums andescape from the valleys with the generalized non-uniform mutation operator subsequently. Tosolve the two problems, a global prediction-based adaptive mutation particle swarmoptimization (GPAM-PSO) is proposed in this paper.The performance of GPAM-PSO is tested on21well-known benchmark problems in30and100dimension separately, compared with9existing PSO in terms of both accuracy andefficiency. The experimental results demonstrate that GPAM-PSO outperforms all referencePSO algorithms on both the solution quality and convergence speed. On the one hand, it canenable all the functions to reach the given accuracy by decreasing the probability to prematureconvergence, and76%（16/21）of the functions get the optimum. On the other hand, it canimprove the efficiency87%more than other algorithms. Also, when it comes to themulti-object problems, GPAM-PSO can perform better than any other compared algorithm.