Um framework de apoio à instanciação de técnicas de seleção de tecnologias de software baseadas em estratégias de busca

Abstract

The quality of software design is directly related to the decisions taken during its execution, because wrong decisions may cause significant damage in a software project. Among the decisions to be performed by a software engineer, an important one would be the selection of technologies to be applied to software projects. Usually, these decisions are made taking into account the experience of the professionals involved in the task. Thus, we limit the exploration of other possibilities that could be most appropriate for such scenario, what could be offered by a scientific approach to support this decision making. This thesis presents a framework for instantiation of software technologies selection techniques by using search-based strategies. For this, the Software Technologies Selection Problem (STSP) was modeled as a combinatorial optimization problem (Minimum Dominanting Set) with the purpose of to attend different and real scenarios of Software Engineering. The proposed framework for the STSP was idealized as a mechanism to support software engineers who are not able to use other generic optimization frameworks in a software project due to tight deadlines and limited resources. It was designed to be integrated with the main meta-heuristic optimization frameworks, such as JMetal and Opt4J, that implement a large number of meta-heuristics. To analyze the feasibility of the proposed modeling and the developed framework, two case studies were conducted in complex and real optimization problems: (i) selection of model-based testing techniques; (ii) selection of requirements elicitation technique for critical embedded systems. The studies were performed using different meta-heuristics and their results indicate their feasibility to support the selection of software technologies.