Skip to Main Content
This paper proposes a boosting algorithm of fuzzy rule-based systems for pattern classification problems. In the proposed algorithm, several fuzzy rule-based classification systems are incrementally constructed from a small number of training patterns. A subset of training patterns for constructing a fuzzy rule-based classification system is chosen according to weights associated to them. The weight for a training pattern is high when it is correctly classified many times. On the other hand, a low weight is assigned to those training patterns that are misclassified many times. Training patterns with a low weight are included in a subset of training patterns for constructing a single fuzzy rule-based classification system. We select the same number of training patterns from each class so that the bias in the number of training patterns among different classes is minimized. In computer simulations, we examine the performance of the boosting algorithm for the fuzzy rule-based classification systems on several real-world pattern classification problems.