基于信息熵的压缩感知流量预测方法研究

Research on Compressed Sensing Traffic Prediction Method Based on Information Entropy

网络流量的预测对控制网络拥塞,提高网络Qos的管理效率以及提高网络利用率有着非常重要的作用。针对目前流量预测大多都是从数学的理论模型入手,从而导致实际流量值与预测精度出现的不一致性,加上蜂窝网络系统用户实际流量变化与基站功率不一致带来网络能量浪费等等,文章提出一种基于时间空间相关的网络流量预测算法。作为压缩感知的先验条件,该算法以网络流量信号的稀疏性与时空相关性相结合,从宏观和微观构造流量矩阵,进行流量矩阵的设置,仿真实验证明该算法精准有效,并运用到实际的网络流量预测中,可以更有效地预测网络流量。

Network traffic prediction plays an important role in controlling network congestion, improving the management efficiency of network Qos, and improving the network utilization rate. The current traffic prediction mostly starts from mathematical models, resulting in the inconsistency between actual traffic values and the predicted values. Besides, network energy is wasted due to the inconsistency between the actual traffic changes of cellular network system users and the base station power. To solve the above two problems, this paper puts forward a kind of time and space related network traffic prediction algorithm. As a prerequisite of compressed sensing, the algorithm combines the sparsity with the temporal and spatial correlation of network traffic signal, establishes and sets the traffic matrix from macroscopic and microscopic aspects. Simulation experiments prove that the algorithm is accurate and effective, and its application to actual network traffic prediction can predict network traffic more effectively.