基于预测模型和自主计算的网络资源管理

为解决综合网络系统中的自主管理问题,以气象水文业务系统为背景,对自主网络管理中的设备状态预测机制和DEN-ng策略模型扩展机制以及结合这两种机制处理网络通信系统中的自主管理问题进行了深入研究,提出了NAR-BP（Nonlinear Auto Regressive-Back Propagation）状态预测模型,并根据观探测设备的测量数据,对该模型进行了实验测试。实验结果表明,NAR-BP状态预测模型的各项误差指标较传统的AR（AutoRegressive）模型算法减少了约4/5。基于DEN-ng策略模型扩展了管理决策组件,使其适应应用层的系统状态预测,通过一个策略驱动的多信道数据通信管理实例展示了DEN-ng扩展策略模型在自主网络管理中应用的可行性。

Network Resource Management Based on Prediction Models and Autonomic Computing

Aiming at the autonomic network management problem in composite functional network system,on the basis of MHBS （ Meteorological and Hydrological Business System） ,in-depth researching of device status prediction and DEN-ng policy model and the combination of above two mechanisms was done. The NAR-BP （ Nonlinear Auto Regressive-Back Propagation） prediction model was proposed and some experiments of the model were done on the basis of real measure data. The result indicated that the SSE （ Sum Squared Error） ,MSE （ Mean Square Error） and MSPE （ Mean Square Percent Error） of NAR-BP prediction model was less than existing linear AR （ Auto Regressive） model about 4 /5. Finally the management decision components was expanded based on DEN-ng policy model for the adaptability of system status prediction on application level providing an instance of policy-driven data communication management with multiple channels to show the feasibility of expansion DEN-ng policy model in autonomic network management.