BIOGEOGRAPHY-BASED OPTIMIZATION FOR COMBINATORIAL PROBLEMS AND COMPLEX SYSTEMS

Biogeography-based optimization (BBO) is a heuristic evolutionary algorithm that has shown good performance on many problems. In this dissertation, three problems are researched for BBO: convergence speed and optimal solution convergence of BBO, BBO application to combinatorial problems, and BBO application to complex systems. The first problem is to analyze BBO from two perspectives: how the components of BBO affect its convergence speed; and the reason that BBO converges to the optimal solution. For the first perspective, which is convergence speed, we analyze the two essential components of BBO – population construction and information sharing. For the second perspective, a mathematical BBO model is built to theoretically prove why BBO is capable of reaching the global optimum for any problem. In the second problem addressed by the dissertation, BBO is applied to combinatorial problems. Our research includes the study of migration, local search, population initialization, and greedy methods for combinatorial problems. We conduct a series of simulations based on four benchmarks, the sizes of which vary from small to extra large. The simulation results indicate that when combined with other techniques, the performance of BBO can be significantly improved. Also, a BBO graphical user interface (GUI) is created for v combinatorial problems, which is an intuitive way to experiment with BBO algorithms, including hybrid BBO algorithms. The third and final problem addressed in this dissertation is the optimization of complex systems. We invent a new algorithm for complex system optimization based on BBO, which is called BBO/complex. Four real world problems are used to test BBO/Complex and compare with other complex system optimization algorithms, and we obtain encouraging results from BBO/Complex. Then, a Markov model is created for BBO/Complex. Simulation results are provided to confirm the model.