Artificial bee colony (ABC) algorithm, which is inspired by the foraging behavior of honey bee swarm, is a novel swarm intelligent optimization algorithm. It shows more effective than genetic algorithm, particle swarm optimization and differential evolution, etc. Due to its simplicity and ease of implementation, ABC algorithm has captured much attention and has been applied to solve many practical optimization problems of diverse domains (such as economics, society and engineering, etc.). However, similar to other evolutionary algorithms, there are some insufficiencies, namely, ABC algorithm is good at exploration but poor at exploitation and its convergence speed is also an issue in some cases.Aiming at the insufficiencies of ABC, it is deeply studied in this thesis. Firstly, inspired by the mutation strategy of improved DEs, and combined the solution search equation of ABCs, several modified solution search equations are proposed. Next, generalized oppositionbased learning method as an improvement measure is introduced to ABC algorithm for increasing its searching ability. In original ABC, while producing a new solution, changing only one parameter of the parent solution results in a slow convergence rate. In order to overcome this issue, the ABC algorithm is modified by adjusting dynamically frequency of the perturbation at each iteration. In addition, a novel calculation to determine and compare the quality of alternative solutions is employed. Based on the techniques, a few improved ABCs for global optimization are proposed. Simulation results on several groups of benchmark functions show that the proposed algorithms improve the performance of ABC at different levels. Finally, one of the proposed algorithms is utilized to solve Periodic Railway Timetable Scheduling Problem. Numerical simulation based on Guangzhou Metro system in China and comparisons with several representative algorithms demonstrate the effectiveness of the proposed approach. In the following, let me explain explicitly what I have done.Chapter 1 is the exordium. In this chapter, a survey to the background and the significance of this thesis is firstly presented. Then, metaheuristic algorithms and their clas sification are introduced, and a brief introduction for several typical metaheuristic algorithms is given. Finally, the organization of this thesis is presented.In Chapter 2, origin and progress of ABC algorithm are introduced, as well as its biological principles, mathematical model and basic procedure are introduced. A comprehensive survey of the advances with ABC and its applications is presented.In Chapter 3, we study improving ABC algorithm and increasing its performance. For the problems of good at exploration but poor at exploitation in ABC algorithm and its poor convergence, some improvement measures are proposed, which include designing new solution search equation, adjusting dynamically frequency of the perturbation at each iteration and modifying the mechanism to determine and compare the quality of alternative solutions, etc. Based on these improvement measures, two modified ABC algorithms are proposed. Experiments are conducted on a set of benchmark functions. The experimental results show that the proposed algorithms are better than, or at least comparable to the original ABC, GABC, and other evolutionary algorithms including DE variants and PSO variants from the literature in terms of convergence performance.In Chapter 4, for the insufficiency in ABC algorithm regarding its search equation, the search strategies in the employed bees phase and the onlookers phase are modified. In the employed bees phase, generalized oppositionbased learning method as a search mechanism is introduced to produce new alternative solutions, and the information of local best solution is utilized to guide the search behavior of bees in the onlookers phase. And then, improved ABC based on local best solution is proposed. The experimental results on a set of 21 benchmark functions show that the proposed algorithm can outperform ABCbased algorithms and other significant evolutionary optimizers in solving complex numerical optimization problems.In Chapter 5, inspired by mutation strategies of existing evolutionary algorithms, two modified solution search equations are firstly proposed. Further, two new designed ABC algorithms (CPABC and OPABC) are given. Then, Periodic Railway Timetable Scheduling Problem (PRTS) is presented, and the model of PRTS problem is modified. Finally, OPABC algorithm is used to solve numerical optimization problems and PRTS problem. Simulation results on a set of 24 benchmark functions show that OPABC algorithm improves the performance of ABC by a large magnitude in solving complex numerical optimization problems and can outperform other evolutionary algorithms being compared. The comparison results on the PRTS problem demonstrate the performance of OPABC is very promising.Chapter 6 is conclusion, in this chapter, the main contributions of this thesis are summarized and expectation for future research are made.
