Artificial Bee Colony Algorithm And Its Applications

How to design an efcient algorithm to solve the problems in various scientifc andengineering felds is essential in science and real applications. Recently, evolutionary al-gorithms (EAs) have shown considerable success in solving optimization problems charac-terized as nonconvex, discontinuous, nondiferentiable, and so on and attracted more andmore attention in recent years. At present, EAs have been broadly applied to diversifedfelds and the research fruits of EAs have already permeated into many disciplines.Artifcial bee colony algorithm (ABC), which belongs to the family of EAs, is basedon simulating the foraging behavior of the honeybee swarm. Due to its simplicity, easeof implementation and few parameters, ABC has captured much attention and has beenapplied to solve many practical optimization problems since its invention. However, atpresent, the research and application on artifcial bee colony algorithm is still in primitivestage. There are still many problems to be studied and solved. For example, similar toother EAs, ABC also faces up to the poor convergence and the low calculating accuracyin solving complex optimization problems. To address these concerning issues, someimproved ABCs for global optimization are proposed in the dissertation for the majorpurpose of increasing the universality, efciency and robustness. The main contributionsand original ideals included in the dissertation are summarized as follows:1. There is still an insufciency in the ABC regarding its solution search equation,which is good at exploration but poor at exploitation. Inspired by diferential evolution(DE), we propose two modifed ABCs (denoted as ABC/best/1and ABC/best/2), whichare based on that each bee searches only around the best solution of the previous iterationin order to improve the exploitation. Experiments are conducted on a set of26benchmarkfunctions. The results demonstrate good performance of ABC/best in solving complexnumerical optimization problems when compared with two ABC based algorithms.2. In order to further improve the exploitation of the algorithm, a novel solutionsearch equation is proposed. In the new search equation, the generated candidate solu-tion is not only around the best solution, but also its search direction is guided by thebest solution, which results in the strong exploitation. Then, to make full use of and bal-ance the exploration of the solution search equation of ABC and the exploitation of theproposed solution search equation, we introduce a selective probability and get the newsearch mechanism (MABC). Experiments are conducted on a set of28benchmark func-tions. The results demonstrate good performance of MABC in solving complex numerical optimization problems when compared with the others algorithms.3. A novel ABC method is presented called as EABC to improve the performanceof ABC. In this method, according to diferent emphases of employed bees and onlookersin the search process, two new search equations for solution update of employed bees andonlookers respectively are proposed to balance exploration and exploitation. Overall, thenew version puts more emphasis on exploiting the domain-specifc knowledge. Finally,simulation results show that EABC is better than, or at least comparable to, other classicor modifed ABC, diferential evolution and particle swarm optimization algorithms fromthe literature in terms of convergence performance for a set of48benchmark functions.EABC also shows promising results for relatively high dimensional problems.4. Inspired by the crossover of genetic algorithm, we frst propose an improvedABC method called as CABC where a modifed search equation is applied to generatea candidate solution to improve the search ability of ABC. Furthermore, we use theorthogonal experimental design (OED) to form an orthogonal learning (OL) strategy forvariant ABCs to discover more useful information from the search experiences. Owing toOED’s good character of sampling a small number of well representative combinations fortesting, the OL strategy can construct a more promising and efcient candidate solution.In this paper, the OL strategy is applied to three versions of ABC, i.e., the standardABC, global-best-guided ABC (GABC), and CABC, which yields OABC, OGABC, andOCABC, respectively. The experimental results on a set of22benchmark functionsdemonstrate the efectiveness and efciency of the modifed search equation and the OLstrategy. The comparisons with some other ABCs and several state-of-the-art algorithmsshow that the proposed algorithms signifcantly improve the performance of ABC.5. To solve the control and synchronization problems of chaotic dynamical systems,we propose an improved ABC (IABC). In IABC, a parameter M is applied to ABC/rand/1and ABC/best/1. Then, in order to take advantage of them and avoid the shortages ofthem, we use a selective probability to control the frequency of introducing ABC/rand/1and ABC/best/1and get a new search mechanism. Numerical simulation based on He′nonMap and comparisons with some typical existing algorithms demonstrate the efectivenessand robustness of the proposed approach.