A modified sensitivity analysis method for driving a multidimensional search in the Artificial Bee Colony algorithm

In this paper, we present an Artificial Bee Colony (ABC) algorithm coupled with a sensitivity analysis method to guide its multidimensional search process. This sensitivity analysis method can evaluate the weight of each dimension of the problem on the objective function computation. We propose a new method for selecting a random neighbor during ABC search phase, using the information of the sensitivity analysis computation. As the algorithm is running, we collect information on the solutions visited and their evaluations. When a sufficient number of evaluation is reached, we launch our sensitivity analysis process to evaluate the influence of each dimension on the objective function result. A weight is then computed on each dimension and promotes the search following these dimensions. The result of this analysis drives the algorithm towards significant dimensions of the search space to improve the discovery of the global optimum.