Research On Hybrid Swarm Intelligence Algorithm And Its Application In Clustering Analysis

The performance of the single swarm intelligence algorithm is more or less limited by the algorithm itself or its principle when solving more complex problems.Generally single swarm intelligence algorithm has the following disadvantages,easy to fall into local extremum,producing premature phenomenon,weak generalization ability and low precision of the result.The hybrid algorithm of swarm intelligence algorithms synthetically considers the differences and complementarities among various swarm intelligence algorithms,and then merges two or more swarm intelligence algorithms together according to some rules to realize the information increment and complementary advantages,so as to enhance the overall optimization performance of the algorithm.The thesis mainly focuses on improving whale optimization algorithm and particle swarm optimization.And then these algorithms are applied to solve the function optimization problems,clustering analysis and fuzzy clustering image segmentation.The main works of this thesis are as follow.1.A whale optimization algorithm with the Levy Flight is proposed.By introducing Levy flight strategy in the process of whale shrinking and encircling prey,the global searching ability and searching accuracy of whale optimization algorithm are improved.Comparing with the WOA and other algorithms on 13 test functions,it is proved that LFWOA can avoid falling into local optimum and has high convergence accuracy.2.Considering the differences and complementarities between PSO and LFWOA,a hybrid swarm intelligent algorithm based on the parallel fusion of Levy flight whale swarm and particle swarm is proposed.The algorithm uses parallel mechanism to realize the collaborative optimization process,and realizes the information interaction among individuals by crossing and replacing operations,so as to increase the diversity of the population in the late iteration period and balance the local exploitation ability and global exploration ability of the algorithm.The simulation results of 23 benchmark functions show that the algorithm has improved in diversity,convergence speed,robustness and optimization accuracy.3.An improved algorithm that combines whale optimization algorithm(WOA)with fuzzy c-means clustering(FCM)algorithm is proposed.The improvement is reflected in four aspects: objective function design,which considers the relationship between intra-class compactness and inter-class separation;adaptive convergence factor,which effectively balances the global exploration and local exploitation ability of whale search process;crossover and mutation strategy,which improves convergence rate and population diversity;monitor mechanism,which realizes dynamic combination between whale algorithm and fuzzy c-means clustering(FCM).By clustering five groups of data sets,it is verified that the clustering results obtained by the hybrid algorithm are compact within the cluster,separated from each other,and have high clustering accuracy and strong robustness.4.To solve the problem about how to select the initial cluster centers,a fast fuzzy c-means clustering(FFCM)method is proposed based on whale optimization algorithm(WOA)for image segmentation.In the method,an improved method is proposed in terms of searching strategy and convergence speed of whale algorithm.In this algorithm,improvement measures are put forward in search strategy and convergence speed of whale algorithm.The simulation results show that the algorithm can reduce the loss of image information,shorten the time of finding the optimal clustering center,and improve the image segmentation effect.