This paper focuses on two essential topics of the fuzzy area. The first is the reduction of fuzzy rule bases. The classical inference methods of fuzzy systems deal with dense rule bases where the universe of discourse is fully covered. By applying sparse or hierarchical rule bases the computational complexity can be decreased. The second subject of the paper is the introduction of some fuzzy rule base identification techniques. In some cases the fuzzy rule base might be given by a human expert, but usually there are only numerical patterns available and an automatic method has to be applied to determine the fuzzy rules. © 2007 Springer-Verlag Berlin Heidelberg.