基于多偏好与变量分解的大规模高维目标优化方法及应用研究

项目名称： 基于多偏好与变量分解的大规模高维目标优化方法及应用研究

项目编号： No.61472366

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 其他

项目作者： 王丽萍

作者单位： 浙江工业大学

项目金额： 82万元

中文摘要： 许多实际应用问题往往同时含有高维目标和大规模变量。针对这类大规模高维目标优化问题，本项目立意将多偏好、变量分解与进化多目标优化相结合，对大规模高维目标优化问题的进化优化方法及其应用进行系统深入的研究。高维目标处理上，构建利用多偏好来同时指导解集排序和维持种群多样性的整体策略；大规模变量处理上，提出面向进化多目标优化的变量分解及协同优化方法；高效进化算法上，以粒子群优化算法为对象，建立不完全依赖于适应度函数值的新个体评价机制及其融合方法；方法应用上，研究新型多目标优化方法在大规模物流网络优化中的应用。本项目的创新之处在于，利用多偏好来同时指导解集排序和维持种群多样性，提出面向大规模多目标优化问题关联变量的识别方法，为大规模物流网络优化等实际应用问题提供新方法。项目的研究将为复杂多目标优化问题的求解提供新的思路，让研究同行对利用多偏好引导和变量分解来解决大规模高维目标优化问题有更深入和更全面的认识。

中文关键词： 进化计算；多目标优化；数据挖掘；复杂网络

英文摘要： Many real-world problems contain both many-objective and large-scale variables. For these large-scale many-objective optimization problems, this project would combine multiple preferences and variable decomposition within Evolutionary Multi-objective Optimization (EMO) to study EMO-based large-scale many-objective optimization method and its application. Firstly, in many-objective handling, the integration strategy that using multiple preferences to guide solution ranking and maintain population diversity is constructed. Secondly, in large-scale variables handling, an EMO-oriented variable decomposition and co-optimization method is proposed. Thirdly, a new individual evaluation method that isn't fully depend on fitness function values is proposed for Particle Swarm Optimizer (PSO). Finally, the new EMO method is applied in heterogeneous network clustering. The main contributions of this research are, utilizing multiple preferences to guide solution ranking and keep population diversity, proposing a variable decomposition for large-scale multi-objective optimization problem and providing new method for solving heterogeneous network clustering and other real-world problems. This research would provide new thinking for solving complex multi-objective optimization problem, and make other researchers have a better understanding on solving large-scale many-objective optimization problems using multiple preferences and variable decomposition.

英文关键词： Evolutionary Computation;Multi-objective Optimization;Data Mining;Complex Network