Improving the Interpretability of Support Vector Machines-based Fuzzy Rules

Abstract: Support vector machines (SVMs) and fuzzy rule systems are functionally equivalent under some conditions. Therefore, the learning algorithms developed in the field of support vector machines can be used to adapt the parameters of fuzzy systems. Extracting fuzzy models from support vector machines has the inherent advantage that the model does not need to determine the number of rules in advance. However, after the support vector machine learning, the complexity is usually high, and interpretability is also impaired. This paper not only proposes a complete framework for extracting interpretable SVM-based fuzzy modeling, but also provides optimization issues of the models. Simulations examples are given to embody the idea of this paper.