基于特征子空间的SOM天基状态检测算法  

SOM Space-Based Resource State Detection Algorithm Based on Feature Subspace

作　　者：刘军 陈锐 肖倩倩 王宇飞 LIU Jun;CHEN Rui;XIAO Qian-qian;WANG Yu-fei(School of Computer Science&Engineering,Northeastern University,Shenyang Liaoning 110169,China)

机构地区：[1]计算机科学与工程学院,辽宁沈阳110169

出　　处：《计算机仿真》2024年第3期51-56,共6页Computer Simulation

基　　金：国家自然科学基金项目(62071134,61671141);中央高校基本科研业务费(N2116015,N2116020)。

摘　　要：在天基信息网中,资源监控系统采集到的性能指标数据量大、冗余特征较多,导致状态检测准确率低、检测时间长等问题。针对以上问题,提出了基于特征子空间的SOM状态检测算法。将特征提取算法嵌入到SOM神经网络模型中,使得网络在训练的同时提取每个类别对应的属性特征,构成状态对应的特征子空间。并利用状态的特征子空间计算特征对于类别的贡献度,优化SOM神经网络的目标函数,进而提高模型对卫星计算任务执行单元状态检测的速度与准确率。仿真了在不同状态检测模型下的检测准确率、检测灵敏度以及检测时间。结果表明,提出的状态检测模型在检测准确率、检测灵敏度以及检测时间等方面都具有较好的性能。In the space-based information network,the resource monitoring system captures a large amount of data and redundant features of performance indicators,which leads to low accuracy of state detection and long detection time.To address this problem,a SOM state detection algorithm based on feature subspace weighting is proposed.The feature extraction algorithm was embedded into the SOM neural network model,so that the network can extract the attribute features corresponding to each category while training,and form the feature subspace corresponding to the state.The contribution of the feature to the category was calculated by using the feature subspace of the state,and the objective function of the SOM neural network was optimized,so as to improve the speed and accuracy of the model for satellite state detection.The experiment simulated the detection accuracy,detection sensitivity and detection time under different state detection models.The results show that the state detection model proposed in this paper has good performance in terms of detection accuracy,detection sensitivity and detection time.

关 键 词：天基信息网 特征子空间 自组织映射神经网络 状态检测 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]