Dissimilarity Computation for Objects of Different Variable Types

Clustering is the process of grouping a set ofphysical or abstract objects into classes of similarobjects is called clustering. A cluster is a collectionof data objects that are similar to one anotherwithin the same cluster and are dissimilar to theobject in other cluster. Measuring the dissimilaritybetween data objects is one of the primary tasks fordistance-based techniques in data mining andmachine learning, e.g., distance-based clusteringand distance-based classification. The quality ofclustering can be accessed based on dissimilaritymeasures of objects which can be computed forvarious types of data. In this paper, we proposegeneral framework for measuring a dissimilaritybetweens various data analysis is proposed. The keyidea is to consider the dissimilarity between twovalues of an attribute as a combination ofdissimilarities between the conditional probabilitydistributions of other attributes given these twovalues. In this system, the similarity is guessed bycomputing the dissimilarity measure between twoobjects. This can get the most similar values andthe least similar values before clustering analysis