Quick Guided Artificial Bee Colony (QGABC) algorithm for training neural networks on classification and prediction tasks

Abstract

The super-metaheurtic are among the most attractive optimization algorithms based on the nature inspired social insects acting like ants, birds and bees. Artificial Bee Colony (ABC) is one of them, which uses the environment behaviors of honey bees for solving different linear and nonlinear complex problems. ABC algorithm has been used for solving a couple of optimization problems efficiently; however, employed and onlooker bees similar actions of inspiring trapped it in local minima and slow convergences speed as well. To increase the performance of a typical ABC with respect to exploration and exploitation process, Guided ABC and Quick ABC algorithms are hybrid named Quick Guided Artificial Bee Colony (QGABC) algorithm. The proposed algorithm has been simulated on Boolean classification and clustering problems through the Neural Networks (NNs) training and testing process. From the simulation results, the QGABC algorithm has outperformed in comparison with the typical ABC, GGABC and QABC algorithms.