The artificial bee colony (ABC) algorithm is a metaheuristic search method inspired by bees' foraging behaviour. With its global search ability in scout bee phase, it can easily escape from local optimum traps in the problem space. Therefore, it is good at exploration. The migrating birds optimization (MBO) algorithm is another recent metaheuristic search method. It simulates birds' V flight formation, which minimizes energy consumption during flight. The MBO algorithm achieves a good convergence to the global optimum by using its own unique benefit mechanism. That is, it has a good exploitation capability. This paper aimed to combine the good exploration property of the ABC algorithm and the good exploitation property of the MBO algorithm via a sequential execution strategy. In the proposed method, firstly, the ABC algorithm runs. This enables solutions to escape from local optimum traps and orientates them to the region in which the global optimum exists. Then the MBO algorithm runs. It performs a good convergence to the global optimum. In the proposed method, some variants of the ABC algorithm and some other well-known optimization algorithms were tested via benchmark functions. It was seen in the experiments that the proposed method gave competitive benchmark test results considering both success rates and convergence performances.

System identification is an important process to investigate and understand the behavior of an unknown system. It aims to establish an interface between the real system and its mathematical representation. Conventional system identification methods generally need differentiable search spaces and they cannot be used for nondifferentiable multimodal search spaces. On the other hand, metaheuristic search algorithms are independent from the search space characteristics and they do not need much knowledge about the real system. The migrating birds optimization algorithm is a recently introduced nature-inspired metaheuristic neighborhood search approach. It simulates the V flight formation of migrating birds, which enables birds to save energy during migration. In this paper, first, a set of comparative performance tests by using benchmark functions are performed on the migrating birds optimization algorithm and some other well-known metaheuristics. The same metaheuristic algorithms are then employed to solve several system identification problems. The results show that the migrating birds optimization algorithm achieves promising optimizations both for benchmark tests and for system identification problems.

No abstract

Migrating birds optimization algorithm (MBO) is a recentlyintroduced nature inspired metaheuristic neighbourhood search approach and simulates V flight formation of migrating birds, which is an effective formation for birds in order to save the energy. Artificial bee colony (ABC) algorithm which is inspired by the bees' foraging behaviour is another powerful optimization algorithm. In this paper, two new variants of MBO algorithm are proposed and a set of performance tests are applied by using benchmark functions. Finally, the proposed methods are employed to train the neural networks which are implemented for nine different data sets in UCI and KEEL web sites. Results show that the proposed methods outperform the original version by performing good convergences to the global optimums. Keywords-Metaheuristic; swarm intelligence; optimization; migrating birds optimization; artificial bee colony optimization I. INTRODUCTIONThe metaheuristics are algorithms that are designed to solve a wide range of optimization problems without having to adapt to each problem in detail. They are generally applied to problems for which there is no satisfactory problem-specific algorithm to solve [1]. Lots of metaheuristic algorithms have been introduced by researchers so far. Some of the first and most popular of them are genetic algorithm [2], simulated annealing algorithm [3], tabu search algorithm [4], ant colony algorithm [5] and particle swarm optimization algorithm [6]. Some recently introduced metaheuristics are differential evolution algorithm [7], harmony search algorithm [8], monkey search algorithm [9], the ABC algorithm [10], firefly algorithm [11], intelligent water drops algorithm [12], cuckoo search algorithm [13-14], bat algorithm [15-16] and the MBO algorithm [17]. Similarly, many researchers worked on the best parameter tunings of these algorithms [18], modified or hybrid forms of them [19-21] and their parallel running methodologies [22][23] in order to get better optimization.Most of these metaheuristics are inspired by the nature and mobile agents interact locally in them. These agents typically explore the search space locally, aided by randomization that increases the diversity of the solutions on a global scale. Thus, there is a fine balance between local intensive exploitation and global exploration to increase optimization performance of the algorithm [24]. We proposed two new variants of MBO algorithm in this paper and it is organized as follows. Section II explains the theoretical backgrounds of the algorithms and