动态多目标协同微粒群优化及其在数据流聚类中的应用

项目名称： 动态多目标协同微粒群优化及其在数据流聚类中的应用

项目编号： No.61473299

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： 张勇

作者单位： 中国矿业大学

项目金额： 80万元

中文摘要： 目标函数多且随环境动态变化的优化问题是非常普遍的。尽管已有进化优化方法可以提高种群跟踪时变Pareto最优解集的能力，但是，这些方法固有的缺陷，使其难以有效处理环境变化不规律、变量维数高的复杂动态多目标优化问题。本项目利用协同进化技术，研究用于动态多目标优化问题的合作型协同微粒群优化理论、方法及其应用。通过研究，拟建立基于环境敏感程度的变量空间划分理论，给出基于完整解集的子种群个体优劣比较策略，提出用于问题求解的合作型协同多目标微粒群优化算法，并将其用于数据流聚类问题。研究成果将为动态多目标优化问题提供一种新的求解途径，提高算法对环境变化的响应速度和求解质量。本项目是自动化、计算机与数学等学科有机交叉、新颖且富有挑战性的研究方向，有非常明确的产业需求，因此，具有重要的理论意义和实际应用价值。

中文关键词： 粒子群优化；协同进化；多目标；动态优化

英文摘要： The multi-objective optimization problems in dynamic environment widely exist in our real life. Although the existing evolutionary algorithms can improve the capability of population to tracke time-varying Pareto optimal set,it is difficult to apply those algorithms in more complicated problems， which have irregular environment and high dimension variables,because of their inherent disadvantages. Based on the co-evolutionary technology,this project studies the cooperative co-evolutionary particle swarm optimization theory and method for solving complicated dynamic multi-objective optimization problems, and their applications. Based on this project, we plan to establish a new dividing theory of variable space based on environment sensitivity, give an individual comparison strategy based on whole solution set, propose a cooperative co-evolutionary multi-objective particle swarm optimization algorithm for solving the above problems, and apply the proposed theory and algorithm in data stream clusering problems. The results of this project should produce a novel solving method for dynamic multi-objective optimization problems,and improve the response speed of algorithm,as well as the quality of solutions. This project is a novel and challenging research orientation with obvious social requirement,which combines automatization,computer science and mathematics.Therefore,it has important theoretical and actual value.

英文关键词： Particle swarm optimization;Co-evolutionary algorithm;Multi-objective;Dynamic optimization