Case-based classification using fuzziness and neural networks

Case-based reasoning may be defined as a model of reasoning that incorporates problem solving, understanding, and learning, and integrates all of them with memory processes. These tasks are performed using some typical situations, called cases, already experienced by the system. There are widespread applications of the concept of case-based reasoning in various decision making processes e.g., medical diagnosis, law interpretation where the knowledge available is usually incomplete and/or evidence is sparse. The article is an attempt in building a case-based pattern recognition system using fuzzy sets and neural networks. Cases are typically labeled patterns which represent different regions or characteristics of the classes. Incorporation of fuzzy set theory helps in selecting the cases from ambiguous/overlapping regions. The methodology is realized in a connectionist framework where the architecture is determined through growing and pruning of nodes, under supervised mode of training, on the basis of fuzzy similarity between patterns