Artificial Bee Colony Algorithm with Improved Explorations for Numerical Function Optimization

A major problem with Artificial Bee Colony (ABC) algorithm is its premature convergence to local optima, which originates from lack of explorative search capability of the algorithm. This paper introduces ABC with Improved Explorations (ABC-IX), a novel algorithm that modifies both the selection and perturbation operations of the basic ABC algorithm in an explorative way. Unlike the basic ABC algorithm, ABC-IX employs a probabilistic, explorative selection scheme based on simulated annealing which can accept both better and worse candidate solutions. ABC-IX also maintains a self-adaptive perturbation rate, separately for each candidate solution, to promote more explorations. ABC-IX is tested on a number of benchmark problems for numerical optimization and compared with several recent variants of ABC. Results show that ABCIX often outperforms the other ABC-variants on most of the problems.