Recently, we introduced a dynamic functional model of the human brain. This model, representing functional connectivity in the brain, is generated from subject-specific physiological data collected using functional magnetic resonance imaging (fMRI). The dynamics of this model are examined using agent-based modeling techniques, wherein a collection of binary agents are embedded as nodes in the network. This model is capable of producing a wide variety of complex behaviors. In this work, we use machine learning techniques to drive the model to produce desired behaviors. The solution space of the model is unreasonably large for a brute-force approach, but we demonstrate that genetic algorithms (GAs) are able to locate optimal model parameters within this space to achieve the desired behavior. We detail the design of a GA specifically suited for this model, and discuss the relevant issues that arise in GA design. Specifically, we explore several fitness functions to accurately quantify the suitability of each potential solution. We examine their strengths and weaknesses, and identify an optimal fitness function for this system. We validate the GA with the optimal fitness function by showing that it can drive the system to produce pre-defined behaviors. The ability of the model to produce pre-defined behaviors indicates that it may be possible to produce physiologically relevant outputs. The model may be very useful for studying the changes in brain dynamics due to neurological diseases or conditions. Additionally, this powerful dynamic brain model may be instrumental in many artificial intelligence settings.