Fractional-order Particle Swarm Optimization And Its Application In Clustering Analysis

Under the background of big data era,the research on swarm intelligence is an important branch of artificial intelligence,and the improvement of swarm intelligence algorithm and its application in data mining have become research hotspots.In the context of big data,due to the characteristics of large data scale,fast update speed and diverse data types,traditional swarm intelligence model gradually fails to meet the requirements of complex problems in the real environment of big data.Therefore,it is of great significance to design new swarm intelligence optimization algorithms for solving complex optimization problems in the big data environment.At the same time,as an important part of data mining technology,clustering analysis has been widely concerned by many scholars at home and abroad,and various clustering algorithms have emerged.As for the clustering of data sets,it can be regarded as an optimization problem,and the optimization objective of this problem is to find the optimal data clustering scheme.Therefore,different optimization methods can be used to improve the quality of clustering analysis.Based on the above research background,this paper selects the most representative algorithm in all swarm intelligence algorithms—particle swarm optimization algorithm(PSO),and the most widely used approach in all clustering analysis methods—partition clustering algorithm as the theoretical basis of this paper.A series of studies are carried out around the relevant algorithms,the specific content is as follows:1)The current research status of particle swarm optimization was analysed,and according to the improvement scheme aiming at the advantages and disadvantages of relevant algorithms,the sigmoid-model self-adaptive fractional order particle swarm optimization algorithm based on swarm activity feedback was designed.This algorithm can according to the current iteration cycle activity value of the population and the active degree of value of each particle,the adaptive adjustment of each particle movement parameters of the next cycle and through the hybrid mutation mechanism to further reduce the algorithm into the risk of premature convergence,greatly improve convergence speed and convergence precision of the algorithm with less computational complexity increased.2)Around the system partition clustering algorithm have great influence on the effect of clustering number and choice of initial clustering center,,according to the characteristics of clustering points designing new particle encoding and optimization rules,a kind of automatic clustering based on fractional particle swarm algorithm combined with the sigmoid-model self-adaptive fractional order particle swarm optimization algorithm was designed,which enables the algorithm to automatically determine the appropriate number of clustering and clustering center location,and improve the clustering quality and robustness of the traditional clustering methods.