Abstract
In this paper we perform a deep investigation about the usefulness of an immune-inspired algorithm to design accurate and compact fuzzy rule bases for classification problems. The algorithm, called Bayesian Artificial Immune System (BAIS), incorporates a mechanism to learn a probability graphical model from the promising solutions found so far. Thus, BAIS utilizes this model to sample new candidate solutions. The probabilistic model utilized here is a Bayesian network due to its capability of expressing the relationships among the variables of the problem, avoiding the disruption of already obtained high-quality partial solutions (building blocks). Besides the capability to identify and manipulate building blocks, the algorithm maintains diversity in the population, performs multimodal optimization and adjusts the size of the population automatically according to the problem. These attributes are generally absent from alternative algorithms, and can be considered useful attributes when generating fuzzy rule bases, thus guiding to high-performance classifiers. BAIS was evaluated in thirteen well-known classification problems and its performance compares favorably with that produced by contenders.