CLUSTER EVALUATION DEVICE, CLUSTER NUMBER CALCULATION DEVICE, CLUSTER DEVICE, CLUSTER EVALUATION METHOD, AND PROGRAM

PROBLEM TO BE SOLVED: To solve a conventional problem of difficulty in supporting appropriate classification of a classification object such as pictures.SOLUTION: A cluster evaluation device comprises: a cluster information storage part which stores two or more cluster information including two or m...

Full description

Saved in:
Bibliographic Details
Main Authors MINO MICHIHIKO, HASHIMOTO ATSUSHI, MORI MIKIHIKO, KASAHARA SHUICHI, INA TAKUO, IIYAMA MASAAKI
Format Patent
LanguageEnglish
Japanese
Published 18.01.2018
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:PROBLEM TO BE SOLVED: To solve a conventional problem of difficulty in supporting appropriate classification of a classification object such as pictures.SOLUTION: A cluster evaluation device comprises: a cluster information storage part which stores two or more cluster information including two or more elements; a calculation part similarity between each element which calculates two or more of similarity between each element included in a cluster, by two or more of each cluster, by using corresponding cluster information; an internal similarity calculation part which calculates internal similarity being similarity representing similarity of the entire inside of each cluster from similarity of two or more of each cluster, by each cluster; a first evaluation value calculation part which calculates a first evaluation value being information regarding a difference between minimum internal similarity being minimum similarity out of the internal similarity and employed internal similarity being one or more internal similarity out of one or more internal similarity other than the minimum internal similarity; and an evaluation value output part which outputs the first evaluation value. The cluster evaluation device can support appropriately classifying a classification object such as pictures.SELECTED DRAWING: Figure 1 【課題】従来は、写真等の分類対象を適切に分類することを支援することが困難であった。【解決手段】2以上の要素を含む2以上のクラスタ情報が格納されるクラスタ情報格納部と、2以上の各クラスタごとに、対応するクラスタ情報を用いて、クラスタに含まれる2以上の各要素間の類似度を算出する要素間類似度算出部と、クラスタごとに2以上の各要素間の類似度から、各クラスタの内部全体の類似度を代表する類似度である内部類似度を算出する内部類似度算出部と、内部類似度のうちの最小の類似度である最小内部類似度と、最小内部類似度を除く1以上の内部類似度のうちの1以上の内部類似度である採用内部類似度との差異に関する情報である第一評価値を算出する第一評価値算出部と、第一評価値を出力する評価値出力部とを具備するクラスタ評価装置により、写真等の分類対象を適切に分類することを支援できる。【選択図】図1
Bibliography:Application Number: JP20160138105