A New Multi-objective Optimization With Preference

The relative high volume of research conducted on evolutionary multi-objective optimization in the last few years. Most of them take Pareto dominance as their selection strategy and do not require any preference information .Most algorithms give a set of solutions but few of them give any analyzing on these solutions which makes it difficult to choose for the decision maker.On the basis of analyzing a large quantity of references home and abroad, the thesis has made a survey on multi-objective optimization with preference. After studying the foundational methods and theories in Attribute Theory, we used Qualitative Mapping (QM) and Attributive Coordinate as multi-objective optimization with preference algorithm, and developed a multi-objective optimization with preference algorithm based on Attribute Theory.On the basis of the algorithm constructed in this article, the process consists of three steps of studying weight, searching project and getting a set of satisfactory solutions. In the first step, some projects are provided to decision makers who will score them, and then we get the preference of the decision makers and curves of the preference of the decision makers. In the second step, every project is expressed by an n-dimensional vector, so the process of getting the solutions is the process of searching the satisfactory vectors in n-dimensional vector space. The searching algorithm is genetic algorithm based on Attribute Theory. On the beginning of algorithm, some projects are generated by random, and then genetic operations such as selection, crossover, and mutation are operated. Then some good projects are survived with high probability. In the last step, using the preference of decision maker, we get the project or a set of projects which are satisfied by decision maker.In order to verify the validity of multi-objective optimization with preference algorithm based on Attribute Theory, we made a lot of experiments on some functions, the result of experiments shows that the method is reasonable and effective. Because of the importance of algorithm on multi-objective optimization with preference, our research is necessary. It provides a new method for...