Adaptive Mutation Rate for the Artificial Bee Colony Algorithm: A Case Study on Benchmark Continuous Optimization Problems

Abstract

A major problem with the Artificial Bee Colony (ABC) algorithm is its premature convergence to the locally optimal points of the search space, which often originates from the lack of explorative search capability of its mutation operator. This paper introduces ABC with Adaptive Mutation Rate (ABC-AMR), a novel algorithm that modifies the basic mutation operation of the original ABC algorithm in an explorative way. The novelty of the proposed algorithm lies in an adaptive mutation strategy that enables ABC-AMR to automatically adjust the mutation rate, separately for each candidate solution of the bee population, in order to customize the degree of explorations and exploitations around each candidate solution, while the original ABC algorithm employs a naïve fixed mutation rate. Besides, a few more explorative schemes and parameter values are employed by ABC-AMR to assist the adaptive mutation procedure. ABC-AMR is evaluated on several benchmark numerical optimization problems and results are compared with the basic ABC algorithm. Results show that ABC-AMR can perform better optimization than the original ABC algorithm on some of the benchmark problems.