高维数据流异常节点动态跟踪仿真研究

Dynamic Tracking Simulation of Abnormal Nodes in High Dimensional Data Flow Timu

针对传统方法难以对异常节点动态实时跟踪,且运行时间较长问题,提出一种基于高维数据流异常节点动态跟踪方法。首先,根据网络数据流的特性,通过时间系列中的平均距离和累计距离,实施高维数据流异常检测。然后建立传感器测量模型得到所有节点的观测方程。再运用最小二乘法,得到目标状态的初始化信息并完成定位。利用网络节点生存期函数,在满足目标跟踪可靠度要求的前提下选取生存期最大的节点参与目标跟踪。采用动态成簇策略,阶段性的选择唤醒任务节点检测目标,根据离散时间线性一致性算法,使其达到可观测状态。最后采用协方差矩阵的平均值来完成高维数据流异常节点动态跟踪。实验结果表明:上述方法能够大幅降低系统的能耗,节省大量时间,具有高效性、准确性和优质的鲁棒性。

It is difficult for traditional methods to track nodes in real time, so a dynamic tracking method for abnormal nodes in high-dimensional data flow was proposed. According to the characteristics of network data flow, the anomaly detection of high-dimensional data flow is implemented through the average distance and cumulative distance in time series was used to detect abnormal nodes in high dimensional data flow. Then, the sensor measurement model was built to obtain the observation equations of all nodes. Moreover, the least square method was used to get the initialization information of target. And then, the location was completed. In addition, the function of network node lifetime was adopted. On the premise of meeting the requirements of target tracking reliability, the node with the largest lifetime was selected to participate in the target tracking. The dynamic clustering strategy was adopted to periodically select the wake-up task and the target of node detection. According to the discrete-time linear consistency algorithm, we could make it observable. Finally, the mean value of covariance matrix was used to dynamically track the abnormal nodes in high-dimensional data flow. Simulation results show that the proposed method can greatlyreduce the energy consumption of system and save significant time. In addition, this method has high efficiency, high accuracy and high robustness.