基于改进小波-Elman神经网络算法的蜂窝网流量预测

摘要

In the wireless network system, theresource dynamic control and energy efficiency improve-ment allclosely depend on the early and precise prediction result of basestation traffic. The characteristic of network traffic volume overspace and time play an important role in traffic prediction. Inthis paper, we make the preprocessing based on both temporaland spatial view for the cellular traffic data generated by alarge population city of China. For each base station, combinethe most similar another one through using k-Nearest Neighbor, determine the most appropriatesliding window size, integrate the Elman Neural Network (ENN)and wavelet transform to achieve the traffic volume prediction.We present numerical results to illustrate the accuracy of wirelesstraffic volume prediction, and we test the performance of ourmethod to demonstrate improvement over some existing methods.%对于在现代蜂窝网资源管理中,动态信道资源和能源效率控制技术的提升,很大程度依赖于早期精准的监测和对蜂窝基站流量的预测.分析基站流量数据,主要通过有效提取基站间隐含的时空信息进行流量预测.在本文中,我们通过对华北某大城市的实测数据,进行了基于时空关联性的分析,采用k-NN算法,获取蜂窝网基站间的时间相关性,选择合适的移动窗口大小,并结合了小波-Elman神经网络(ENN)算法来实现流量预测.最后,通过量化蜂窝网流量预测的准确度,并与先前存在的其他方法进行对比,得出了本文提出的方法有优越性.