PERFORMANCE EVALUATION OF MULTIVIEWPOINT-BASED SIMILARITY MEASURE FOR DATA CLUSTERING

Abstract: Some cluster relationship has to be considered for all clustering methods surrounded by the data objects which will be applied on. There may be a similarity between a pair of objects which can be defined as a choice of explicitly or implicitly. We in this paper introduce a novel multiviewpoint based similarity measure and two related clustering methods. The main distinctness of our concept with a traditional dissimilarity/similarity measure is that the aforementioned dissimilarity/similarity exercises only a single view point for which it is the base and where as the mentioned Clustering with Multiviewpoint-Based Similarity Measure uses many different viewpoints that are objects and are assumed to not be in the same cluster with two objects being measured. By utilizing multiple viewpoints, countless descriptive evaluation could be accomplished. In order to assist this declaration, the theoretical analysis and empirical study are carried. Depending on this new measure two criterion functions are proposed for document clustering. We examine them with certain distinguished clustering algorithms which use other preferred coincident measures on different group of documents in order to verify the improvement of our scheme