Sensitivity Analysis of Bayesian Networks in COTS-based Software

Abstract

The process of developing software applications by integrating one or more Commercial Off-The-Shelf (COTS) components has received much attention lately because it provides potential benefits including shortening the development time, reducing effort and shrinking budgets as well as improving the quality of the final product. However, COTS-based development (hereafter CBD) in particular the evaluation and selection of COTS components, which is an essential activity in CBD, is not a trivial task and associated with various challenges. One of the most critical challenges is uncertainty inherent to COTS-related information and their vendors. Ignoring the uncertainty challenge negatively influences the quality of COTS selection decisions. In this paper, a bayesian-based evaluation model  is extended to allow the allocation of various weights to evaluation criteria. We also investigate the impact of using various weights on the belief about the satisfaction level for various COTS candidates. Furthermore, the paper shows how the analytic hierarchy process (hereafter AHP) is used along with the model to rank various candidates. A digital library system is selected as an example to illustrate how the model along with AHP help decision makers to select the most promising COTS candidate