Multi-objective Optimization Algorithm And Preference Multi-objective Decision Making Based On Artificial Immune System

Nowadays, the research on the biological information system has been one of the hottest and the most important topics in the field of artificial intelligence. Biological immune system is a highly evolved, parallel, and distributed adaptive system. Its information processing ability provides important insights into the fields of computer science, which area is called artificial immune system. Meanwhile, many real-world problems are influenced and determined by multiple factors, and the corresponding area is called multi-objective optimization, which has also received wide concerns from the related scholars. Based on the artificial immune system, this thesis discusses the multi-objective optimization algorithms and the preference multi-objective decision making. The main works in this thesis are as follows:1) Based on the simulated annealing strategy, this thesis combines artificial immune system and multi-objective optimization together, and proposes a simulated annealing based immunodominace algorithm called SAIA. SAIA adopts some novel operators, such as simulated annealing based adaptive hyper-mutation operator, simulated annealing selection operator, and population pruning operator. The experimental study proves that SAIA performs well in terms of maintaining population diversity and converging to the Pareto fronts.2) This thesis proposes a novel dominance relationship called region-dominance, and based on the region-dominance the thesis proposes a hybrid multi-objective immune algorithm with region-dominance called HMIAR. From the experimental study, it can be found that HMIAR can incorporate the decision maker's preference information into the multi-objective optimization problems and help the decision maker to obtain a set of well distributed and evolved solutions.3) For well solving the dynamic multi-objective optimization problems, this thesis proposes a novel dominance relationship named sphere-dominance, and proposes a sphere-dominance preference algorithm called SPIA. SPIA adopts two mutation types which are Predicted mutation and Gaussian mutation. It can be concluded from the experimental study that SPIA performs excellent when solving the dynamic multi-objective optimization problems with the preference information.The work was supported by the National High Technology Research and Development Program (863 Program) of China (No.2009AA12Z210), the National Natural Science Foundation of China (No.60803098), and the Program for Cheung Kong Scholars and Innovative Research Team in University (No.IRT0645).