Ensembling Classification Systems by a Fuzzy Rule-Based System

Abstract

In this chapter, we propose an ensembling method for pattern classification problems. In our ensembling method, two different types of fuzzy rule-based systems are used. One is for classifying input patterns. We refer to this type of fuzzy rule-based systems as fuzzy rule-based classification systems. The other type of fuzzy rule-based systems determines which classification systems are used for the classification. This type of fuzzy rule-based systems is referred to as fuzzy rule-based ensembling systems. Our ensembling method consists of one fuzzy rule-based ensembling system, several fuzzy rule-based classification systems, and a gating node that is used for final classification. An input pattern is presented to both types of fuzzy rule-based system. Each of the fuzzy rule-based classification systems determines which class the input pattern is from, and the fuzzy rule-based ensembling system assigns a credit of the classification to each fuzzy rule-based classification system. In the gating node, all the information is collected and the final classification of the input pattern is performed. In computer simulations, we examine the performance of our ensemble learning method on several real-world pattern classification problems. Simulation results show that the performance of our ensembling method is better than the best single fuzzy rule-based classification system. We also show the simulation results of our method on unseen patterns to see how well our method generalizes.KeywordsFuzzy RuleInput PatternEnsembling MethodTraining PatternPattern Classification ProblemThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.