In decision approaches applied to network control, it is common to use a function that encompasses a set of parameters
weighted through empirically defined values. However, these weights are usually not optimized, and the error from the optimal
value will be propagated to the decision function. This paper presents a method to determine the best input weight parameters according to a predefined payoff function using a genetic algorithm (GA). We have extended the GA with respect to its key elements, e.g. chromosomes coding scheme and fitness function.
The method is quite general and suits any simulation-based optimization problem with multiple discrete input parameters, without requiring the whole system behavior to be expressed into a mathematical formula. The optimization method is tested in our approach for social-aware nodes’ association in mobile
networks, and the results show that it autonomously searches for the input weight values that lead to the best solution for the association decision, improving the results when compared to empirically defined weights.