earticle

영어

As Internet has been wildly spreaded and it's technique is advanced, the use of computers has been routinized and almost data are stored in computers. Accordingly, many companies and researchers have tried to find the relations in these tremendous data and the one way is to use clustering algorithm which is used to find out similar data set in the entire data set and to discover the common properties. In early period, clustering algorithm was performed based on a main memory of a computer and PAM(Partitioning Around Medoids) was representative, which can be complemented k-means algorithm defeat. PAM performs clustering by using the medoid of data instead of means. PAM works well in small data set but it is difficult to apply it to large data set. Therefore, CLARA(Clutering LARge Application) shows up to be used in large data set. This algorithm samples data from large data set and applies PAM to the sample data. CLARA has limits caused by the fixed samples in each clustering stage and has a problem that if the good mediod is not sampled then the result of the clustering becomes not good. CLARANS(Clustering Large Application based upon Randomized Search) overcomes these problems by drawing a sample with some randomness. This algorithm executes clustering using k mediod set extracted in the processing of clustering in each stage. The main objective is to compare and analyze the algorithms which are popularly used for the clustering of big data.