퍼지 서포트 벡터 기계 분류-그룹 소속 가능성 = Fuzzy Support Vector Machine Classification - Possibility detail pertaining to group data

다국어 초록 (Multilingual Abstract)

Support vector machine, which is an universal approximator based on the Vapnik's statistical learning theory, is very popular to many scien-tists for its unique learning method and good generalization perfor-mance. Instead of using empirical risk minimization which is generally used by statistical learning method such as multi-layer perceptron, radial basis function network, etc. support vector machine uses struc- tural risk minimization to reduce its generalization error rate to find the appropriate objective function. Also, like other general learning method.
support vector machine can be applied to the classification problem (pattern recognition) and function estimation (nonlinear regression).
The result of the experiment showed that it was advanced to SVM classification of crisp data. In reality, however, there are many kinds of inaccurate and ambiguous fuzzydata. This kind of research has not been conducted inside South Korea before now. My research has shown that fuzzy input data can be handled using this fuzzy support vector machine classification (FSVC) algorithm.
The basic idea of FSVC algorithm is to transform an existing weight vector into a fuzzy weight vector. The fuzzy weight vector norm is defined as the sum of the centernorm, (where the center has the maxi-mum membership function value), and spread norm, (where the spread norm indicates both left end point and right end point). In this research I have proposed a new fuzzy spread plane method that differs from the original SVM classification hyperplane. This research enables us to make efficient decisions with much more accurate information whist illustrating more possibility detail pertaining to group data.

목차 (Table of Contents)

• Ⅰ. 서론
• Ⅱ. 퍼지 서포트 벡터 기계 분류
• 1. 비퍼지 자료
• 2. 퍼지 자료
• Ⅲ. 그룹 소속 확률
• Ⅵ. 실증사례
• Ⅴ. 결론