Use of fuzzy if-then rules for pattern classification

An efficient fuzzy partitioning method of a feature space for pattern classification problems is proposed in this article. A feature space is initially decomposed into some overlapping subspaces depending on the relative positions of the pattern classes found in the training samples. To reflect the pattern classes by the generated subspaces, a few fuzzy if-then rules are then obtained in terms of a relational matrix. The relational matrix is utilized in the modified compositional rule of inference in order to recognize an unknown pattern. The proposed system is capable of handling incomplete and other imprecise information both in the learning and processing phases. The effectiveness of the system is demonstrated on two real life problems. The proposed system is capable of reflecting the nonoverlapping, overlapping and no-class regions of the feature space by providing output decisions in terms of single, multiple and null choices. The multivalued outputs are found to be superior than existing classical and fuzzy approaches