Uma abordagem para o problema do escalonamento dinâmico em projeto de software

Abstract

The Software Project Scheduling Problem consists of allocating employees to tasks during a software development. A good schedule allows the project not only to reduce costs and its makespan, but to minimize the impacts that the project could suffer when exposed to dynamic events of the real world, such as the exit of an employee or even the emergence of new requisits. This is a combinatorial problem, and and as it is a complex task to find good solutions, search techniques may be applied to solve the problem. This work investigates the adequation of the application of the meta-heuristic Nondominated Sorting Chemical Reaction Optimization to the Dynamic Software Project Scheduling Problem. An adaptation to the problem was proposed and validated through experiments. The results were compared with the ones of the metaheuristis Nondominated Sorting Genetic Algorithm II. The analysis showed that the Nondominated Sorting Chemical Reaction Optimization produced better outputs than the Nondominated Sorting Genetic Algorithm II in projects with few employees and tasks, despite its efficiency decreased and became worse as the project increased.