| Project/Area Number |
09838016
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
感性工学
|
|---|
| Research Institution | Nagoya University |
|---|
Principal Investigator |
FURUHASHI Takeshi Dept.of Information Electronics, Nagoya University Associate Prof., 工学研究科, 助教授 (60209187)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KODAMA Tetsuji Dept.of Information Electronics, Assistant Prof., 工学研究科, 助手 (50262861)
ISHIGURO Akio Dept.of Computational Science and Eng., Associate Prof., 工学研究科, 助教授 (90232280)
UCHIKAWA Yoshiki Dept.of Computational Science and Eng., Prof., 工学研究科, 教授 (20023260)
|
|---|
| Project Period (FY) |
1997 – 1998
|
| Project Status |
Completed (Fiscal Year 1998)
|
| Budget Amount *help |
¥2,500,000 (Direct Cost: ¥2,500,000)
Fiscal Year 1998: ¥1,200,000 (Direct Cost: ¥1,200,000)
Fiscal Year 1997: ¥1,300,000 (Direct Cost: ¥1,300,000)
|
|---|
| Keywords | fuzzy modeling / fuzzy neural network / genetic algorithm / fuzzy inference / knowledge acquisition / knowledge discovery / 知識獲得 / 免疫システム / 記号推論 |
|---|
| Research Abstract |
This research project was aiming at developing methods for modeling of human Kansei. This project was carried out in three approaches : (a) modeling using fuzzy neural network, (b) modeling using genetic algorithm, (c) modeling using immune system. Main results were obtained in (a) and (b) approaches.
These results are summarized as follows :
(a)Modeling using fuzzy neural network
Two of the most important steps in modeling of Kansei are : (1) selection of apppropriate combination of input variables which greatly influence the outputs, and (2) appropriate division of input space which is spanned by the selected input variables. This research project developed a method for the selection of input variables using fuzzy neural networks and genetic algorithm. This project also developed two methods for the division of input space. One put emphasis on comprehensibility of identified fuzzy rules. The other was to obtain concise models. The developed method could identify balanced models with precision and conciseness.
(b)Modeling using genetic algorithm
This research project developed a method using genetic algorithm for identification of fuzzy rules from data unevenly distributed in the input space. The problems in this type of fuzzy modeling were that the genetic algorithm needed much computation time and the acquired fuzzy rules were difficult to understand. The combination of input variables and the shapes of membership functions varied in each fuzzy rule. This research project developed a method for acceleration of rule identification and a method for clarification of identified fuzzy rules. The developed methods could extract acquired rules by minimizing the degradation of precision of fuzzy rules.
|
