DANTAS, EMANUEL; http://lattes.cnpq.br/6639945575396326; DANTAS FILHO, Emanuel.
Resumo:
Risk management processes are present in traditional and agile methodologies. Risks are
abstract and subjective events, which makes management a complex activity. Several techniques are applied to reduce subjectivity and help manage risk in software projects. These techniques address subjectivity directly or indirectly, and among the most commons are the probability and impact matrix, checklists, brainstorming, and Artificial Intelligence-based methods. State of the art in risk management for software projects involving technological innovation presented gaps in the treatment of the investigated approaches. In particular, understanding risks in software projects, their identification, and proper treatment must consider the technological characteristics of a project, significantly when aspects of innovation are associated. This is one of the challenges faced by the research community and is a critical success factor in any software design and development methodology. Given this gap, this study proposes an approach to risk management in software projects based on Bayesian Networks, focusing on identifying and monitoring risks. The objective of the approach is to reduce subjectivity in risk management activities using factors related to the projects’ technological characteristics. A qualitative study was carried out with 25 professionals from ten development companies to elicit risk factors associated with the technological characteristics of software projects. Then, through a Knowledge Engineering process, a Bayesian Network was built. The validation focused on innovation projects. Initially, through simulation scenarios that represent real cases of software projects, a static validation was performed. In each scenario, eight professionals from a technological innovation organization evaluated the actions recommended by the approach. Finally, through an experiment, a dynamic validation was performed. Evidence corresponding to real-world data from six of the organization’s projects was used as input to the Bayesian Network to assess the network’s predictions and the understanding of the project professionals. Based on the results, the proposed approach can assist in decision-making in identifying and monitoring risks.