基于重要性分数的CUR矩阵分解识别WSN异常节点  

CUR matrix factorization based on importance scores recognizing WSN abnormal nodes

作　　者：谢丽霞[1] 田宇祺 XLE Li-xia;TIAN Yu-qi(School of Computer Science and Technology,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300

出　　处：《计算机工程与设计》2024年第4期997-1003,共7页Computer Engineering and Design

基　　金：国家自然科学基金民航联合研究基金项目(U1833107)。

摘　　要：现有异常节点识别方法普遍存在精度低或者功耗大等不足,针对这些不足,提出一种基于低秩矩阵分解的无线传感器网络异常节点识别算法。根据传感器节点的特性进行特征选取和属性矩阵构建;通过改进的CUR矩阵分解方法计算异常矩阵;依据异常矩阵中节点向量的总体均值和标准差设定控制限,判断节点是否发生异常。实验结果表明,与其它异常识别方法相比,该方法具有较高的识别准确率。The existing abnormal node recognition methods generally have low accuracy or high power consumption.In view of these deficiencies,an abnormal nodes recognition algorithm of wireless sensor network based on low-rank matrix factorization was proposed.The feature selection and attribute matrix were constructed according to the characteristics of the sensor node.The node calculated the exception matrix using the improved CUR matrix factorization method.The control limit was set by determining the overall mean and standard deviation of the node vector according to the exception matrix.Experimental results show that the present method has high recognition accuracy compared with other abnormity recognition methods.

关 键 词：矩阵分解 属性矩阵 异常节点 重要性分数 休哈特控制图 相关性程度 异常矩阵 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]