應用歸納學習策略由數值資料產生模糊規則

Abstract

In this paper, we have attempted to apply the concept of fuzzy sets to machine learning. We assume, first, that the classification into positive and negative examples in the training set is a degree of positiveness and negativeness between 0 and 1;second, that an inexact matching between a concept description and an example is allowed; and third, that data may contain wrong, uncertain, and linguistic information. A new inductive learning problem is then formulated so as to induce a concept description that covers almost all of positive examples and almost none of negative ones. Therefore, a fuzzy learning algorithm by the version-space strategy is proposed to manage wrong, uncertain, and linguistic information under imprecision and noise environments