特征变量的灰色模糊动态聚类选择方法  

Feature Variable Selection Method for Model Based on Fuzzy Clustering and Grey Correlation

作　　者：赵民全[1] 白如顺 李杰然[1] 

机构地区：[1]中国人民解放军92785部队

出　　处：《雷达科学与技术》2016年第3期305-310,共6页Radar Science and Technology

摘　　要：为了在保证计算精度的前提下使模型简化并便于计算,要尽量少地用对模型影响较大、相互独立的特征变量进行建模。在进行特征变量选择时,既要考虑选择对主因素有重要影响的变量,也要排除各影响变量间的多重相关性的干扰。首先建立各特征变量同费用的灰色关联度,根据关联度的大小对众多特征变量进行排序,排除关联度相对极小并同其他因素关联度差异明显的特征变量,减少次要影响因素对估算结果造成的干扰;其次,应用基于特征权值的模糊动态聚类方法,并突出近期数据的重要性,对影响因素进行聚类分析,排除影响因素间多重相关性的干扰。论文通过实例分析进行了验证。In order to simplify the cost model under the premise of ensuring the accuracy,several small main factors is used to establish model.For characteristic variable selection,the variables which have important implications on the cost no high multiple correlation each other should be selected.According to grey correlation between facters,the method of fuzzy clustering based on feature weight is proposed to establish cost model.It highlights the importance of recent factor,cluster analysis of factors,and excludes interference of multiple correlation.The proposed method is verified by an example.

关 键 词：灰色关联分析 模糊聚类 变量选择 特征加权 

分 类 号：TN957[电子电信—信号与信息处理]