Fuzzy classification based on pattern projections analysis

摘要

A method of measuring the comparative efficiency of features and building decision rules in the problem of fuzzy pattern recognition by features is proposed. The method is based on the analysis of the structure of the training set's binary shadows composition on co-ordinate hyperplanes in description space. A number of computer runs were performed to examine the behaviour of the proposed criterion while changing the size of the training set and the mutual disposition of fuzzy set classes in the description space. In all the experiments the classes that take part in recognition process were simulated by fuzzy sets with Gaussian membership function. In addition, some experiments were performed to determine the reliability of a decision rule constructed by the proposed method. The dependence of the extent of the object's recognition on the size of the training set and the mutual disposition of classes in the description space were examined. The experimental results have indicated the efficiency of the proposed criterion application in the problem of fuzzy pattern recognition by its features. Rules for fuzzy pattern classification are proposed that use a space of features.