Research On K-medoids Clustering Optimization Algorithm Based On Swarm Intelligence

With the rapid development of network and information technology,how to extract useful information from the massive data has become one of hotpots for many scholars.data mining is emerge under this background.Clustering analysis is a very important branch of data mining that can be classify different types of massive data sets,and it is widely used in many fields.Based on the existing problems of the traditional clustering algorithm,this paper proposed separately an improved algorithms which combined glowworm swarm optimization(GSO)and an improved which combined algorithms bacterial foraging(BFO)with traditional clustering algorithm by analyze and study the traditional clustering algorithm.The main research of this paper is as follows:(1)Aiming at the traditional K-medoids clustering algorithm has sensitive to the initial cluster centers,poor stability and easy converge to a non-global optimum value defects.This paper proposed a K-medoids algorithm based on GSO which using manifold distance instead of Euclidean distance as a measure method and using GSO to optimize the initial cluster centers(K-medoids clustering algorithm based on glowworm swarm optimization,GSO-medoids).The optimization algorithm GSO-medoids take advantage of the swarm intelligence GSO with the advantages of good stability and flexibility to overcome the traditional method only able to analyze spherical data,poor clustering effect and other defects.Experiments show that GSO-medoids algorithm has higher accuracy and good robustness..(2)Aiming at the traditional K-medoids clustering algorithm has sensitive to the initial cluster centers and the clustering effect is not good for high dimensional data defects.This paper proposed a K-medoids algorithm based on BFO(K-medoids clustering algorithm based on bacterial foraging optimization algorithm,BFO-medoids).The improved BFO-medoids algorithm makes full use of the swarm intelligence BFO has strong search ability and stronger global optimization capability,which is on the basis of the advantages of the original K-medoids algorithm,significantly improved the clustering performance of the algorithm,and suitable for high-dimensional data sets.