Optimization of Agent-Based Models: Scaling Methods and Heuristic Algorithms

Questions concerning how one can influence an agent-based model in order to best achieve some specific goal are optimization problems. In many models, the number of possible control inputs is too large to be enumerated by computers; hence methods must be developed in order to find solutions that do not require a search of the entire solution space. Model reduction techniques are introduced and a statistical measure for model similarity is proposed. Heuristic methods can be effective in solving multi-objective optimization problems. A framework for model reduction and heuristic optimization is applied to two representative models, indicating its applicability to a wide range of agent-based models. Results from data analysis, model reduction, and algorithm performance are assessed.