ABSTRACT The Component-Based Software Development achieves software reusability by creating software systems from the current software rather than building from scratch. Successful reuse requires high quality components in a repository with suitable description, classification and retrieval mechanism. The component selection process consists of picking up a component from the given set based on some quality attributes. Using the fuzzy clustering techniques, the similar components are grouped together for choosing the best candidate. The main goal of this paper is to propose a select the most optimal components from the retrieved set of components. Fuzzy clustering plays a vital role in software component selection. This paper proposes a fuzzy relation based fuzzy clustering technique with the following advantages: i) It eliminates the need of choosing a particular similarity function. As choosing the similarity measure was very crucial in fuzzy clustering also the results may vary for different similarity measures even for the same set of data. ii) With the proposed technique multi-dimensional data numerical as well as categorical can be handled. iii) The prior specification of the number of clusters is not required. iv) The components are clustered based on multi-features rather than considering one or two features. The algorithm is validated on a case study.
