Research On Clustering Algorithm Based On Improved Simplified Particle Swarm Optimization

Cluster analysis,as a research hotspot,plays an important role in the fields of statistics,biology,information retrieval,pattern recognition and machine learning.K-means algorithm as one of the classical clustering algorithm has some advantages such as simple principle,easy to implement,can explain the strong wait for an advantage,but there are clustering number it is difficult to determine,is sensitive to the initial center,shortcomings and so on easy to fall into local optimal point,make it difficult to get the global optimal solution in the algorithm,different size of different categories of data or non-convex data gathered the effect not beautiful.In addition,swarm intelligence optimization algorithm is the result of human learning and exploration from some group behaviors in the biological world,and clustering is often regarded as a special optimization problem,so particle swarm optimization algorithm can be used to solve the clustering problem.This paper improves the simplified particle swarm optimization algorithm and applies it to the clustering algorithm to improve the performance of the clustering algorithm.To this end,two k-means clustering algorithms based on the improved simplified particle swarm optimization are proposed:(1)An improved simplified mean particle swarm optimization K-means clustering algorithm(ISMPSO-AKM)is proposed.On the one hand,on the basis of simplified particle swarm optimization,the neighborhood optimal particle is added to improve the position formula by linear combination of individual optimal position,global optimal position and neighborhood optimal position.On the other hand,an inertia weight based on cosine function and logarithmic function is constructed to realize dynamic adjustment of inertia weight.In addition,AKM clustering algorithm is introduced to determine the number of clusters and dynamically obtain the initial center,which further improves the accuracy of the algorithm.(2)This paper proposes a hybrid algorithm based on simplified particle swarm optimization and k-means clustering.This algorithm integrates the grouping idea of the lion swarm algorithm into the simplified particle swarm optimization algorithm,and divides the particles into three groups for optimization.Each group uses different learning factors and learning dimension vectors to help the population to perform different search mechanisms,thus enhancing the diversity of the population.In addition,using Darwin's competition law of survival of the fittest as reference,the introduction of population breeding is conducive to the particle jumping out of the local optimal position,and the global search performance of the algorithm is improved.Finally,it is combined with k-means clustering algorithm.