Um método baseado em modelo probabilístico para auxílio na detecção de problemas na utilização do scrum em projetos de desenvolvimento de software.

Abstract

There is a high rate of software development projects that fail. Whenever problems can be detected ahead of time, projects may have better chances of success, and therefore save money and time. Recently, the usage of agile methodologies and frameworks have been increasing due to its lightweight practices, rules and principles that allow frequent changes during project execution. Scrum is the most popular agile framework. Even though it is composed of simple practices, rules and principles, professionals find it hard to apply it in the industry, specially, in teams used to traditional project management. To help ScrumMasters to fulfill their duty of facilitating the usage of Scrum, this dissertation presents a method to detect problems regarding its application in software development projects. The method is cyclic and uses a probabilistic model to present project data to the ScrumMaster. Given its capacity to handle uncertainties and flexibility to modifications, Bayesian Networks were used to implement the model. The model was validated using scenarios to test it and the method through a case study in two projects in a company. The results show that using the method helps to detect problem in software development projects using Scrum with a positive cost-benefit and useful to guide the team to achieve excellence.