Uma abordagem baseada em riscos de software para seleção de requisitos

• Ano de defesa: 2017
• Autor(a) principal: Amaral, Aruan Galves Nascimento
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Informática; Programa de Pós-Graduação em Informática; UFPB
• Palavras-chave em Português: Problema do próximo release; Otimização multi-objetiva; Gerenciamento de riscos; Requisitos de software; Next release problem; Multi-objective optimization; Risk management; Software requirements; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/12908

In incremental development approaches, there is a great interest to deliver system releases in a timely manner, while avoiding budget overruns and maximizing the perceived satisfaction for the stakeholders involved in the project. Thus, the requirements selection process is a key-factor in identifying a good or optimal subset of candidate requirements that meet these conditions. Traditional techniques that involve manual processes for selecting and prioritizing requirements have limitations when addressing a large number of requirements. In such a direction, the Next Release Problem (NRP) presents a computational model for this decision process, evolving from a simple single-objective approach, with a maximum allowed budget, to multi-objective approaches that make the decision process more flexible, without restricting the model to pre-fixed limits related to the goals to be achieved on the release. Despite this evolution, most of the contributions for this problem does not address software risks, which is a key-factor that may deeply impact on project cost and stakeholders’ satisfaction. Therefore, this dissertation proposes a risk-based approach for a multi-objective next-release problem, in which risks are incorporated into the cost and satisfaction evaluation for the system to be delivered. In order to validate such a proposal, besides a systematic review comparing the current literature related to the proposal’s theme, this dissertation presents results of three distinct metaheuristics capable of dealing with multiple objectives to be optimized. These algorithms are applied to two semi-real datasets that reveal the efficiency and potential applicability of the proposed approach.