摘要
针对卫星网络业务具有自相似的特点,介绍了一种基于集合经验模式分解的业务组合预测方法(EEMD)。该方法利用EEMD的分解特性,将具有自相似的网络流量分解成多个只具备短相关性的本征模态函数(IMF),这样便可使用传统的流量预测算法进行预测。文中使用人工神经网络与自回归滑动平均模型(ARMA)这两种方法进行预测。最后将多个本征模态函数(IMF)预测的结果相加作为原始信号的预测结果,实验证明此方法有更高的精度。为了迎合卫星实时性的需求,给出了硬件的框架,该框架采用DSP与FPGA相结合的构架实现连续数据的EEMD实时处理。
For the self-similarity of satellite network traffic, a novel business portfolio prediction method based on EEMD (Ensemble Empirical Mode Decomposition) is proposed. This method uses the decompo- sition characteristic of EEMD to divide self-similar network traffic into multiple IMF ( Intrinsic Mode Func- tion) components merely with short dependence, so that traditional methods could be applied to the predic- tion. In the experiment, artificial neural networks and ARMA model are used to do the prediction. Final- ly, the summation of all IMFs perdition data is taken as the prediction result of original signal, and experi- ment verifies that this method could enjoy fairly high accuracy. In order to cater for real-time requirement of satellite network, the hardware framework is also proposed. This framework uses a combined architec- ture of DSP and FPGA to achieve EEMD real-time processing of continuous data.
出处
《通信技术》
2015年第11期1285-1289,共5页
Communications Technology
基金
国家自然科学基金(No.91338201)~~