摘要
为了解决制造系统状态描述的问题,需要从众多的属性中择优选择合适的属性集合,以便降低属性的冗余度,减少计算量.提出了用免疫离散粒子群算法进行属性选择的方法,给出了属性选择的粒子表达、适应度函数的定义以及免疫机制.通过仿真实验给出了描述制造系统状态的入选属性集合,并进行了对比实验,将待选属性集合、入选属性集合和落选属性集合作为支持向量机的输入,来比较3种情况下分类的准确性和验证属性选择的有效性.实验结果表明,经过选择后的属性集合分类准确性大大高于另外两种情况,从而实现对制造系统状态的有效识别,为在不同的状态下采取合适的调度规则建立了基础.
A large number of properties need to be selected into the appropriate attributes sets to describe the manufacturing system state in order to reduce the redundancy of attributes and reduce the computational complexity. A feature selection method was proposed based on immune binary partial swarm optimization. Particle expression of feature selection, definition of fitness function and immune mechanisms were given. The selection results were presented through numerical simulation. A comparative experimentation was designed to testify the effectiveness of the method. Three attribute sets: candidate sets, selected sets and deselected sets were used as the input of support vector machine respectively. Results demonstrate that the classification accuracy of selected sets is greatly better than the other. Then the system state can be effectively recognized and the appropriate rule can be adopted for the corresponding manufacturing system state.
出处
《浙江大学学报(工学版)》
EI
CAS
CSCD
北大核心
2009年第12期2203-2207,共5页
Journal of Zhejiang University:Engineering Science
基金
国家自然科学基金资助项目(50675201)
浙江省科技计划资助项目(2006C11237)
关键词
离散粒子群算法
生物免疫
支持向量机
特征选择
binary partial swarm optimization (IBPSO)
biology immune
support vector machine (SVM)
feature selection