The Algorithm Of Soft Subspace Clustering Based On Particle Swarm Optimization

As an important research topic in the field of clustering algorithm of data mining,subspace clustering algorithm has great effect on the processing procedure of high-dimensional data,sample points distance and clustering validity etc.At the same time,it is more accurate to combine the subspace with a better optimization algorithm,so that the algorithm solves the global convergence clustering center.It has a wide range of applications in text analysis,machine learning,information retrieval,and other fields.However,the subspace clustering algorithm still exists many.problems in practical applications because of local convergence,being sensitive to initial value of clustering center.The objective function and subspace search strategy decide the performance of soft subspace clustering,and cluster validity analysis is the main indicator of its performance.In this paper,a soft subspace clustering algorithm based on particle swarm optimization is proposed,which could automatically give cluster number.Firstly,a novel fuzzy weighting soft subspace objective function is designed by k means-type clustering framework,which combines inter-cluster separation with feature weight.Secondly,in order to jump out of the local best,particle swarm optimization with inertia weight(?PSO)is used as a global search strategy.Finally,the optimal cluster number is selected by the proposed cluster validity function.The experimental results demonstrate that SC-?PSO improves the clustering accuracy and automatically determines the optimal cluster number.