The quantification of stakeholders plays a fundamental role in the selection of appropriate requirements, as their judgement is a significant criterion, as not all stakeholders are equally important. The original proposals modelled stakeholder importance using a weighting approach that may not capture all the dimensions of stakeholder importance. Furthermore, actual projects involve a multitude of stakeholders, making it difficult to consider and compute all their weights. These facts lead us to search for strategies to adequately assess the importance concept, reducing the elicitation effort. We propose grouping strategies as a means of reducing the number of stakeholders to manage in requirement selection while maintaining adequate stakeholder coverage (how selection meets stakeholder demands). Our approach is based on the salience of stakeholders, defined in terms of their power, legitimacy, and urgency. Diverse strategies are applied to select important stakeholder groups. We use k-means, k-medoids, and hierarchical clustering, after deciding the number of clusters based on validation indices. Each technique found a different group of important stakeholders. The number of stakeholder groups suggested experimentally (3 or 4) coincides with those indicated by the literature as definitive, dominant, dependent, and dangerous for 4 groups; or critical, major, and minor for 3 groups. Either for all the stakeholders and for each important group, several requirements selection optimisation problems are solved. The tests do not find significant differences in coverage when important stakeholders are filtered using clustering, regardless of the technique and number of groups, with a reduction between 66.32% and 87.75% in the number of stakeholders considered. Applying clustering methods to data obtained from a project is useful in identifying the group of important stakeholders. The number of suggested groups matches the stakeholders’ theory, and the stakeholder coverage values are kept in the requirement selection.

Selecting the appropriate requirements to develop in the next release of an open market software product under evolution, is a compulsory step of each software development project. This selection should be done by maximizing stakeholders’ satisfaction and minimizing development costs, while keeping constraints. In this work we investigate what is the requirements interactions impact when searching for solutions of the bi-objective Next Release Problem. In one hand, these interactions are explicitly included in two algorithms: a branch and bound algorithm and an estimation of distribution algorithm (EDA). And on the other, we study the performance of these not previously used solving approaches by applying them in several instances of small, medium and large size data sets. We find that interactions inclusion do enhance the search and when time restrictions exists, as in the case of the bi-objective Next Release Problem, EDAs have proven to be stable and reliable locating a large number of solutions on the reference Pareto front.

Requirement engineering is a key issue in the development of a software project. Like any other development activity it is not without risks. This work is about the empirical study of risks of requirements by applying machine learning techniques, specifically Bayesian networks classifiers. We have defined several models to predict the risk level for a given requirement using three dataset that collect metrics taken from the requirement specifications of different projects. The classification accuracy of the Bayesian models obtained is evaluated and compared using several classification performance measures. The results of the experiments show that the Bayesians networks allow obtaining valid predictors. Specifically, a tree augmented network structure shows a competitive experimental performance in all datasets. Besides, the relations established between the variables collected to determine the level of risk in a requirement, match with those set by requirement engineers. We show that Bayesian networks are valid tools for the automation of risks assessment in requirement engineering.

Directed graphic models based on conditional independence provide a compact and concise representation of an expert's subjective belief about existing relationships between variables. Faced with the task of building a greater model, each expert must be a specialist in some subset of the whole knowledge domain. It would be desirable to aggregate the knowledge provided by those specialists under the form of graphical models into a single and more general representation. This article studies the consensus model that would be obtained by combining two graphs associated with Bayesian networks and applying the union and intersection of their independencies. © 2003 Wiley Periodicals, Inc.