Use of derived heuristics in improved performance of evolutionary optimization: An application to gold processing plant

The importance of using heuristics in an optimization algorithm is well established, particularly in solving complex real-world problems. It is then expected that users know certain key problem information a priori and are able to implement the information in a suitable optimization algorithm. However, in many problems, such problem information may not be available before an optimization task is performed, thereby making the heuristics-based algorithms difficult to be implemented. In this paper, we suggest a ‘derived heuristics’ based optimization methodology for this purpose. In such a method, past results from an optimization algorithm are utilized to derive problem heuristics and then used in a future applications to achieve a faster and more accurate optimization task. Heuristics can also be derived from the optimization run and used in subsequent iterations. In a particular gold processing plant optimization problem, we demonstrate the use of derived heuristics by developing a customized evolutionary optimization procedure which is capable of handling various complexities offered by the problem, in a way which is much better than a classical point-based method and a population-based generic approach. The results of this paper is motivating for evolutionary computation researchers to apply the methodology to other more complex real-world problems.