Comparison of heuristic algorithms with performance metrics on benchmark functions

In today's information age, complex problems can be solved in a much shorter time with the increased use of information and communication technologies. Artificial intelligence systems that can think and develop like a human and optimization techniques that try to find the best are being developed with technological advances. The heuristic algorithms which are the techniques used in the optimization process are developed by inspiration from the natural life of the creatures, and reach quickly and easily to the solution that is closest to the best solution. Heuristic algorithms are available in many types and new ones are being developed day by day. In this study, Artificial Bee Colony (ABC), Biogeography-Based Optimization (BBO), Differential Evolution Algorithm (DE), Cuckoo Search Optimization (CSO), Imperialist Competitive Algorithm (ICA), Particle Swarm Optimization (PSO) among many swarm-based heuristic algorithms are used reasons such as widely used in the scientific world, be open source code, and their achievements are better than others. Each of these selected algorithms was run 30 times in 2, 5 and 10 dimensional search spaces with the same initial positions and conditions to find global minimum point on the 8 benchmark functions frequently used in the literature. As a result of runs, the performances of the algorithms were evaluated based on the results of performance metrics such as the best cost, worst cost, accuracy, stability, time and standard deviation. Comparisons of the algorithms according to the cumulative mean performance values showed that the best performance was given by DE, followed by PSO. The CSO, which gives the closest results to the DE in terms of stability and accuracy, is best in terms of running time and is the third in the cumulative average performance evaluation. In the cumulative mean performance evaluation, the ICA was fourth, while the BBO, which gave the closest results to each function, was fifth. ABC, which produces far metric values from the other algorithms, is the last one in the performance evaluation used in this study and it shows the lowest cumulative mean performance value.