Application of a genetic algorithm to optimize quality assurance in software development

Abstract

Quality Assurance is an important aspect of the software development lifecycle. With declining Department of Defense dollars, the development of a Quality Assurance scheme, that minimizes total software development project costs in large scale systems, is extremely valuable. This research aims at developing such a scheme which will provide a staffing profile for Quality Assurance. As there are no analytical solutions available to solve this nonlinear optimization problem and the potential search space of all possible solutions is extremely large, a genetic algorithm is used to arrive at an optimal solution. The results indicate that the solution obtained using this approach performs better than several other approaches, such as expert simulators and pattern search techniques, that have been attempted. The scheme is developed using a software project simulation model that incorporates data from an actual software development project. The simulation model allows for the examination of the consequence of staffing profile decisions on total project cost.