Validation of Software Component Selection Algorithms based on Clustering

Abstract

Objectives:To search the components that provides desired functionality, from the finite set of component set by the use of software component selection process. The selection process helps in choosing the optimal set of components from the third party repository. Methods/Statistical Analysis:In order to select the optimal set of component having multiple attributes, clustering is found to most suitable technique as revealed in the literature. This paper presents the validation of clustering based algorithms used for software component selection. It mainly covers the fuzzy-c means clustering and subtractive clustering. It also includes the earlier software component selection techniques proposed by the authors, hybrid XOR based clustering technique and fuzzy relation based fuzzy clustering. Findings:The Fuzzy c-means technique requires the need of mentioning the number of clusters centers in advanceand the radii of the cluster in case of Subtractive clustering are required. The disadvantage of Hybrid XOR based clustering is its dependency on subjective judgment of the developer. The FREFCOSCO algorithm has eliminated the usage of similarity index. It is able to deal with multi-attributes component and can generate the optimal set of components. Application/Improvements:The algorithms are validated on a set of components taken from an online repository. The improvement in the FREFCOSCO algorithm can be done by using an appropriate validity mechanism.