Abstract: Aiming at the polybasic and heterogeneous characteristics of uncertainty industrial data,a rough set attribute reduction algorithm was proposed.The attribute kernel parameter of rough set was taken as immune bacterins to encode the antibody.A niche immune memory was introduced in clonal selection process to optimize with sharing mechanism coordinately,which could overcome the blindness of crossover and mutation operators,improve the diversity and stability of antibody distribution and promote the preservation of excellent individual in original antibody group and memory base.By optimizing the approximate standard of attribute classification,the ability of excellent individual to get clone expansion was improved and the opportunity of affinity maturation was realized.Simulation experimental result illustrated that the approach was an effective and quick way in solving industry data attribute reduction.

Key words: rough set, immune sharing, clone selection, niche, attribute reduction

摘要： 针对不确定性工业数据多元、异类的特性,提出一种基于免疫记忆机制的粗糙集属性约简方法。将粗糙集属性核参数作为疫苗对抗体进行编码。在免疫克隆选择过程中引入小生境免疫记忆与共享机制进行协同优化,克服进化策略中交叉和变异算子操作的盲目性,提高抗体群分布的多样性及稳定性,促进原始抗体群及记忆库中优良个体的保存。将属性的分类近似标准作为适应度目标进行优化,增强优秀个体获得克隆扩增、实现亲和力成熟的机会。实验仿真及工业故障诊断对比实验结果表明,该算法对于工业数据属性决策约简快速且有效。

关键词: 粗糙集, 免疫共享, 克隆选择, 小生境, 属性约简