基于免疫聚类的特征数据浓缩方法

针对诊断特征数据中重复或相似事例样本和特征参量之间可能存在的相关性，提出一种有效的特征数据双向压缩预处理方法，该法在不损失数据隐含的特征知识的前提下，能有效降低学习机器的学习负担．在进行样本参量的降维处理时，基于主元分析的思想，采用一种改进的主元分析(MPCA)方法用于横向数据压缩，在压缩样本数量时，综述和比较了现有的各种聚类算法，借鉴生物体自然免疫系统中克隆选择以及免疫网络自稳定等有关机理，提出了基于主元核相似度的免疫聚类算法用于纵向数据压缩．仿真实验验证了所提方法的有效性．

Feature Data Enriching Approach Based on Immune Clustering

Aiming at the relativity between repeated or similar samples and characteristic parameters during diagnosis of characteristic data, an effective data analysis approach for characteristic data compression from bi direction is presented, which can reduce the burden of learning machine without losing the connotative characteristic knowledge of characteristic data. At the first step of the algorithm, based on the theory of principal component analysis, a modified principal component analysis (MPCA) approach is adopted to reduce the dimension of data horizontally, then after comparing existing clustering algorithms, an immune clustering algorithm is put forward based on similarity measurement of principle component core for vertical reduction by using related mechanism of clone selection as well as immune network self stabilization in natural organic immune system for reference. Finally, its effectiveness is proved by the simulation experiments.