Generation of Management Rules through System Dynamics and Evolutionary Computation

Decision making has been traditionally based on a managers experience. This paper, however, discusses how a software project simulator based on System Dynamics and Evolutionary Computation can be combined to obtain management rules. The purpose is to provide accurate decision rules to help project managers to make decisions at any time in the software development life cycle. To do so, a database from which management rules are generated is obtained using a software project simulator based on system dynamics. We then find approximate optimal management rules using an evolutionary algorithm which implements a novel method for encoding the individuals, i.e., management rules to be searched by the algorithm. The resulting management rules of our method are also compared with the ones obtained by another algorithm called C4.5. Results show that our evolutionary approach produces better management decision rules regarding quality and understandability.