Design of transparent mamdani fuzzy inference systems

In this paper, we propose a technique to design Fuzzy Inference Systems (FIS) of Mamdani type with transparency constraints. The technique is based on our Crisp Double Clustering algorithm, which is able to discover transparent fuzzy relations that can be directly translated into a human understandable rule base. As a key feature, the user can tune the granularity level of the rule base so as to properly balance the trade off between accuracy and transparency. The resulting FIS bears a transparent knowledge base that can be easily understood by human users and can be effectively used to solve soft computing problems. The work is accompanied by an illustrative example that show the validity of the approach.