论文部分内容阅读
针对网络流量预测,提出一类基于自组织映射(self-organizing map,SOM)神经网络的局部自回归(auto-regressive,AR)方法.根据SOM的联想记忆在时域的推广,在矢量量化临时联想记忆(vector-quantized temporal association memory,VQTAM)建模技术的基础上,给出具有多个局部线性AR模型的AR-SOM方法,基于前K个获胜神经元用权值代替输入向量建立单一时变局部AR模型的K-SOM方法,以及在完成数据向量聚类的同时,更新多个局部AR模型系数的LLM(local linear map)-SOM方法.相对于全局模型,基于SOM神经网络的局部AR方法能够灵活给出有效的监督神经结构,降低了计算复杂度.将本文方法应用于不同的网络流量预测实例中,并与现有方法相比,实验结果表明所提出的方法能有效地改善预测精度,且性能更好.
Aiming at the network traffic prediction, a local auto-regressive (AR) method based on self-organizing map (SOM) neural network is proposed.According to the generalization of SOM associative memory in the time domain, Based on the vector-quantized temporal association memory (VQTAM) modeling technique, an AR-SOM method with multiple local linear AR models is given. Based on the fact that the first K winning neurons use the weights instead of the input vector to establish a single time K-SOM method to change local AR model, and LLM (local linear map) -SOM method to update multiple local AR model coefficients while completing data vector clustering.Compared with the global model, local AR based on SOM neural network The method can flexibly provide an effective supervised neural structure and reduce the computational complexity. The proposed method is applied to different network traffic prediction examples. Compared with the existing methods, the experimental results show that the proposed method can effectively improve the prediction Accuracy, and better performance.