A fast artificial bee colony algorithm variant for continuous global optimization problems

Since its creation in 2005 by D. Karaboga the ABC algorithm proved to be very effective in approaching a wide variety of research optimization problems. However, some drawbacks were also experienced related mainly to a poor exploitation capability (which makes the algorithm relatively slow) and poor success rates when highly non-linear optimization problems with unstructured modes are approached. In order to improve the performance of the ABC algorithm, in both efficiency and success rate, the paper presents a set of proposed enhancements to the original ABC algorithm. The novel proposed ABC variant, Fast ABC (F-ABC), was tested against two known variants of ABC, the original algorithm proposed by D. Karaboga ([1]), and an improved variant, Gbest-guided Artificial Bee Colony (GABC) ([2]). The testing was conducted by employing an original testing methodology over a set of 11 scalable, multimodal, continuous optimization functions (10 unconstrained and 1 constrained) with known global solutions. The novel F-ABC algorithm clearly outperformed the older variants in both efficiency and success rate over the test functions which present unstructured modes, while for the remaining test functions the results were mixed.