Multi-objective Optimization for Mixed-integer Programs through Hybrid Genetic Algorithm with Value Function Modeled by Neural Networks

With the aim of developing a flexible optimization method for managing conflict resolution, this paper concerns itself with multi-objective mixed-integer programs. For this purpose, we have proposed an intelligence supported approach that combines genetic algorithm with mathematical programs (hybrid genetic algorithm) to derive the best-compromise solution. Also we have developed a novel modeling method of value function using neural networks, and incorporated it into the approach which employs a simulated repair operation of DNA. As a result, we can provide a practical and effective method in which the hybrid strategy maintains its advantages of relying on good matches between the solution methods and the problem properties such as a genetic algorithm for unconstrained combinatorial optimization and a mathematical program for constrained continuous ones. Finally, by taking an example for site location problems of hazardous wastes, we have examined the effectiveness of the proposed approach numerically.