Entropy based statistical mechanical fuzzy clustering method and its visualization
| Project/Area Number |
19500201
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Sensitivity informatics/Soft computing
|
|---|
| Research Institution | Gifu National College of Technology |
|---|
Principal Investigator |
YASUDA Makoto Gifu National College of Technology, 電気情報工学科, 教授 (80353275)
|
|---|
| Project Period (FY) |
2007 – 2009
|
| Project Status |
Completed (Fiscal Year 2009)
|
| Budget Amount *help |
¥3,120,000 (Direct Cost: ¥2,400,000、Indirect Cost: ¥720,000)
Fiscal Year 2009: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2008: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2007: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
|
|---|
| Keywords | ファジィ理論 / 情報統計力学 / ファジィクラスタリング / ファジィc平均法 / ファジィエントロピー / 確定的アニーリング法 / シミュレーテッドアニーリング法 / Tsallisエントロピー / 可視化 / 自己組織化マップ / 遺伝的アルゴリズム / 競合共進化アルゴリズム / フォーカシング問題 |
|---|
| Research Abstract |
The fuzzy clustering methods which combine the deterministic annealing (DA) approach with fuzzy entropy or Tsallis entropy are examined. By maximizing fuzzy entropy or Tsallis entropy within the framework of the fuzzy c-means (FCM) method, membership functions similar to the statistical mechanical distribution functions are obtained. The characteristics of these entropies and membership functions are investigated from the statistical mechanical point of view.
Both fuzzy entropy and Tsallis entropy based FCM are formulated as DA clustering. Numerical experiments showed that these algorithms can divide data more properly and stably than the standard clustering method. The very fast annealing (VFA) method is applied to DA for cooling schedule of the temperature. Also, the membership functions are approximated by linear functions. The obtained results indicate that the proposed method can shorten computational times.
Multidimensional questionnaire data were clustered and, by omitting some dimensions, the resultant clusters and memberships were visualized.
|
Report
(4 results)
Research Products
(25 results)
