Title:
Systems and methods for determining membership values for training data for fuzzy neural networks
Document Type and Number:
Japanese Patent JP6362991
Kind Code:
B2
Abstract:
PROBLEM TO BE SOLVED: To provide a system for determining the membership value of fuzzy neutral network learning data.SOLUTION: The system includes: a data set acquisition unit 101 for acquiring a plurality of learning data used in the learning of a fuzzy neutral network, each including a combination of the value of each of a plurality of explanatory variables and the value of an objective variable; a rank classification unit 102 for classifying the learning data into a plurality of ranks on the basis of the magnitude of the value of the objective variable; a frequency calculation unit 103 for finding, for each rank, frequency distribution for each of the plurality of explanatory variables, standardizing the maximum of frequency in the frequency distribution to 1, and finding the standardized frequency of each of the plurality of explanatory variables; and a membership value assignment unit 104 for defining the average value of the standardized frequency of values of the plurality of explanatory variables in each of the plurality of learning data as the membership value of each of the plurality of learning data.SELECTED DRAWING: Figure 5
Inventors:
Daisaku Kimura
Kenichi Yamagata
Arai study
Kenichi Yamagata
Arai study
Application Number:
JP2014213101A
Publication Date:
July 25, 2018
Filing Date:
October 17, 2014
Export Citation:
Assignee:
Azbil Co., Ltd.
International Classes:
G06N3/04; G05B23/02; G06N3/08; G06N7/02
Domestic Patent References:
| JP8314881A |
Attorney, Agent or Firm:
Yoshiyuki Inaba
Toshifumi Onuki
Akihiko Eguchi
Kazuhiko Naito
Toshifumi Onuki
Akihiko Eguchi
Kazuhiko Naito
Previous Patent: Roll storage box
Next Patent: LOW RESIDUAL STRESS TYPE PIPE BODY CONNECTING STRUCTURE
Next Patent: LOW RESIDUAL STRESS TYPE PIPE BODY CONNECTING STRUCTURE