%0 Journal Article %T 基于灰支持向量回归机预测适应值的交互式集合进化计算 %T Set-based interactive evolutionary computation with forecasting fitness by grey support vector regression %A 郭广颂 %A 文振华 %A 郝国生 %A GUO,Guang song %A WEN,Zhen hua %A HAO,Guo sheng %J 控制与决策 %J Control and Decision %@ 1001-0920 %V 35 %N 2 %D 2020 %P 309-318 %K 灰支持向量回归机;隐式性能指标;交互;适应值预测;集合进化 %K grey support vector regression(GSVR);tacit indices;interactive;forecasting fitness;set-based evolution %X 个体适应值的高精度预测和高效的进化策略对于提高进化优化算法性能至关重要.针对现有大规模种群交互式进化计算个体适应值估计误差较大以及传统进化策略搜索效率较低的问题,提出一种基于灰支持向量回归机的个体适应值预测方法和大规模种群集合进化策略.建立基于灰支持向量回归机的适应值预测模型,给出4种集合进化个体比较测度,同时提出新的集合进化个体自适应交叉和变异概率.基于上述策略,采用NSGA-II范式设计一种交互式集合进化优化算法.将该算法应用于RGB颜色One-max优化问题,以表明所提出个体适应值预测方法和集合进化策略的有效性. %X Efficient forecasting fitness and evolutionary strategies are more critical for improving evolutionary algorithm optimization performance. Most current interactive evolutionary computation with large population has larger fitness estimation error and lower efficiency adopting the traditional evolutionary strategy. Accordingly, a fitness prediction method based on grey support vector regression and a set-based evolutionary strategy are proposed. Then, four comparative measures of set-based individuals evolution are defined, and adaptive crossover and mutation probability are proposed. Based on above strategies, a set-based interactive evolutionary algorithm is designed by using the powerful NSGA-II algorithm for optimizing tacit indices problems. The proposed algorithm is applied to RGB color One-max optimization problem, and its outstanding performance is experimentally demonstrated. %R 10.13195/j.kzyjc.2018.0729 %U http://kzyjc.alljournals.cn/kzyjc/home %1 JIS Version 3.0.0