基于偏好信息的多目标微粒群优化算法研究

Study on multiobjective particle swarm optimization algorithm based on preference

在实际决策过程中,决策者可能并不需要完全获悉所有的决策方案,而是只对一些特定方案产生兴趣,对此,提出指定目标间重要关系和给定目标空间参考点情况下的多目标微粒群优化算法.以格栅作为解的多样性保持策略,对于给定目标间重要关系的偏好信息,可以获得特定区域的多个解;对于给定参考点的偏好信息,可以同时获得多个特定区域中的多个解,有利于决策者进行更有效的决策.通过对典型测试问题的仿真实验,验证了本算法的正确性和有效性.

During practical making decision, the maker need not know all the solutions to the problem, but is interested in certain solutions. To meet this requirement, multi-objective particle swarm optimization algorithm based on importance relationship among the objectives and reference points in objective space is proposed, which employs grid strategy to keep solutions diversity. More than one solutions located in certain area can be got when importance relationship among the objectives is specified. And more than one solutions located in more than one certain areas can be also obtained when preference based upon reference points is indicated., which is beneficial to the decision maker making efficient and reliable decisions. Simulation results of a series of classical benchmark problems show the correctness and effectiveness of the algorithm.