METHOD FOR FAST CLUSTERING USING SCALAR OPERATION

摘要

PURPOSE: A method for the fast clustering using a scalar operation is provided to process the clustering fast by applying each scalar operation to the similarity of a sum of square distance and the similarity of the derivation variance of the vector cosine and the vector inner product while performing the hierarchical clustering for the high dimensional and vast data. CONSTITUTION: The input data is read(101) and the read data allots to a new cluster(102). In case that a number of the clusters is larger than a set maximum value, the nearest two clusters are incorporated. The similarity between the incorporated cluster and the existent cluster is calculated by not a vector operation but the scalar operation(105). If the input data is read completely, the clustering is stopped. In case that the number of the clusters is not larger than the set maximum value, the previous processes are repeatedly carried out.