Evidence Based Information Fusion in Intuitionistic Fuzzy Pattern Recognition

In many practical domains, the pattern recognition involves uncertainty in identification and impreciseness in the patterns. Encountering these twoissues provides us more accurate solution which hasrobustness. On the other hand, the Intuitionistic Fuzzy Sets (IFS) as a generalization of the regular Fuzzy Sets,provides a convenient framework to model both uncertainty and impreciseness. Also, the Evidence Theory represents applicable approaches for Information Fusionusing combinatorial rules. In this paper, we proposed an approach for pattern recognition in whichfeature existence in the sample is depicted through IFSand leads to a feature vector constructed of membership and non-membership functions for each feature. Takinginto account several observations regarding a sample feature vector, they are fused using an evidence combinatorial rule. Having obtained a single fused feature vector of the sample, the similarity between the Intuitionistic Fuzzy Sets of uncertain pattern feature vectorsand samples are represented using an IFS similaritymeasure. To examine this approach in practice, we applied it on a medical diagnosis problem. Our experimentshowed that this approach could satisfactorily classify the unknown imprecise samples correctly, so as it yields rational and acceptable results.

Keywords:

Intuitionistic Fuzzy Sets , Evidence Theory , Information Fusion , Pattern Recognition , Medical Diagnosis