Research On Clustering Algorithm Based On Particle Computing

Clustering is a collection of data objects in accordance with the process of sample similarity divided into multiple classes.Clustering analysis is unsupervised learning, so it has the ability of independent learning.Clustering the data space is the purpose of all objects in the class were divided into different clusters.A close object is divided into the same class, a remote object into different clusters.Clustering analysis as a kind of data pretreatment process is often used in many applications,it is further on the basis of the data analysis and data processing.People with the most reasonable and effective way to clustering analysis of data space, and then to get to know and explore the internal relation between the things.Moreover, it can also be applied to some analysis algorithm of pre-processing steps such as association rules.Now, clustering analysis has been widely applied in various fields, such as weather analysis, image processing and so on the need for the field of data processing.With the continuous improvement of modern science and technology level,the rapid development of network,and continuously reform and innovation of computer technology,large quantities of data constantly emerging.How to extract meaningful from these data valuable information become the people have been concerned.With various new problems appear constantly,for the handling of the data by cluster analysis technology ability of the demand is higher and higher,the technology will face a new challenge after another.Only by constantly the ability to improve the clustering algorithm of the data,to solve these new problems.As a result, all kinds of new clustering algorithm arises at the historic moment. Of course also include clustering algorithm based on partitioning, these algorithms in practical application has achieved expected effect.Due to the real world information is imprecise, uncertain, incomplete, ambiguous, which makes the application of uncertainty reasoning has become more important.And because of the granular computing accords with the objective law of human for problem solving, and able to cover the fuzzy set and rough set theory, and many other fields, therefore in to deal with fuzzy and uncertain knowledge, and the knowledge reasoning, granular computing has very strong advantages and potential.Based on the clustering analysis, clustering algorithm and granular computing theory, on the basis of the research status at home and abroad,this paper introduces the definition of clustering and type of clustering algorithm.The basic principles of k-means clustering algorithm and K-medoids clustering algorithm are introduced in detail,their performance are analyzed.The scholars in recent years the study of clustering algorithm based on dividing present situation has carried on the comb, its specific application instance has made a brief introduction.The definition of particle,level,structure, etc are given.The basic problems of granular computing and the three main granule computing model are introduced,focus on clustering granularity principle in detail,clustering algorithm is given based on mechanical analysis of principle of thinking framework. Features and main work of this article is:puts forward two new optimal initial center center of K-medoids clustering algorithms, and gives the specific ideas and algorithm, and through artificial simulated data set and UCI machine learning database standard data set on the simulation experiment, prove their effectiveness.