The component-based software engineering (CBSE) consists of component selection, qualification, adaptation, assembly and updating of components according to the requirements. The focus of this paper is software component selection only. Now-a-days many selection processes, techniques and algorithms are proposed for this task. This paper presents generalized software component selection architecture using clustering. The architecture is divided into four tiers namely Component Requirements and Component Selection Tier, Query and Decision Tier, Application logic tier with Clustering and Component Cluster Tier. The architecture offers manifold advantages like i) presenting a generalized architecture where the existing techniques can be applied, reducing the search space for the component selection. ii) It also illustrates the usage of clustering in the software component selection without the need for pre- specification of number of clusters and considering more than two features while clustering. iii)The cluster validation is performed to check the correctness of the clusters. This complete selection process is validated on a representative instance of set of components.

To achieve successful reusability of components a disciplined development approach is required which is the component based software engineering(CBSE).The software component selection is a vital part of this approach. It consists of defining an evaluation criteria based on user requirements and depending on this the repository is searched and shortlisted components are presented to the user. Due to availability of large number of components offering same type of functionality it is difficult to select a particular component based on available description. This paper presents a multiobjective optimization model for component selection purpose and solves it using preemptive goal programming approach by using an optimization tool LINDO. Subsequently, an illustrative case study is given where the components are taken from an online repository and goal programming is applied for getting the most optimal component. However, this model is applicable when the repository is small but for larger set of components it needs to be validated.