Prediction function for NSGA2

The prediction function needs to "play" a couple more times with the environment and choose the best action, based on the best reward that was returned. This is necessary because it needs to consider the starting point of the environment (state).