DATA CLASSIFIER, DATA CLASSIFICATION METHOD, DATA CLASSIFICATION PROGRAM, AND RECORDING MEDIUM

摘要

PROBLEM TO BE SOLVED: To provide a technology for classifying the group of distribution data. SOLUTION: This data classification method includes: a step (S5) for summarizing the data of the histogram of distribution data belonging to each cluster as a histogram representing each of a plurality of clusters acquired by preliminarily classifying the group of distribution data showing the distribution of the group of vector data in a multi-dimensional space; a step (S6) for calculating symmetric KL information quantity as similarity between distribution data based on the histogram of the distribution data belonging to each cluster and the histogram representing each cluster; a step (S7) for updating the cluster by reclassifying the distribution data to any cluster based on symmetrical KL information quantity; a step (S8) for judging whether or not the assignment destination of the distribution data has been changed before and after the update of the cluster; and a step (S9) for outputting a distribution data classification list when the assignment destinations of all the distribution data are not changed. COPYRIGHT: (C)2008,JPO&INPIT