Classification and clustering of information objects based on fuzzy neighborhood system

Supervised and unsupervised classification given a family of fuzzy neighborhood on a set of information objects to be retrieved is considered. An information object implies any type of objects to be retrieved, e.g., documents, keywords, images, and Web pages. We do not distinguish between terms and documents as in traditional setting of the vector space model. Instead, information link is used and the concept of fuzzy neighborhood is introduced. Classification rules based on the nearest neighbor, K nearest neighbor, and fuzzy K nearest neighbor are proposed. Agglomerative clustering algorithms are moreover developed on the basis of similarity measures defined on the neighborhood. Illustrative examples are given.