Software component clustering and classification using novel similarity measure

The similarity measures such as Euclidean, Jaccard, Cosine, Manhattan etc present in the literature only consider the count of the features but does not consider the feature distribution and the degree of commonality. There is a significant research carried out for designing new similarity measures which can accurately find the similarity between any two software components. The distribution of component features in the software components has important contribution in evaluating their degree of similarity. This is the key idea for the design of the proposed measure. The main objective of this research is to first design an efficient similarity measure which essentially considers the distribution of the features over the entire input. We then cany out the analysis for worst case, average case and best case situations. The proposed measure is Gaussian based and preserves the properties of Gaussian function and can be used for clustering and classification of software components. (C) 2015 The Authors. Published by Elsevier Ltd.