HIGHLY ACCURATE DATA CLASSIFICATION APPARATUS UTILIZING SPHERICAL SELF-ORGANIZATION MAP, AND PROGRAM FOR CREATING HIGHLY ACCURATE DATA CLASSIFICATION APPARATUS

摘要

PROBLEM TO BE SOLVED: To display the whole spherical surface by respective colors included in data groups by allocating a color code to which the data group belongs to each node based on positional relationship between the spherical surface positions of learning-ended labeled data displayed on a spherical self-organization map and respective unlabeled nodes.SOLUTION: The spherical self-organization map corresponding to a plurality of reference vectors is created and code book vectors of the map are stored. Then, nodes and input data are stuck together to a spherical surface to generate a tree diagram, and while storing the tree diagram for labeled data, a distance of the labeled data is calculated so as to find out which label out of respective labels in labeled data is closest to the node concerned and a label sort held in the closest label is allocated to the node. A sort of each node is distinguished as the result, thereby each node is classified by color in accordance with the sort.