摘要
网络并发式流量特征具有信号时间可预测性,通过对网络流量的解卷积测度特征提取,提高对网络流量的预测性能。传统法方法采用粒子群优化算法实现对网络流量的特征测度盲解卷积分析,对原始信号的统计信息提取效果不好。提出一种基于粒子群退化重采样的网络流量解卷积测度提取算法,构建并发式网络流量序列采集模型,设计粒子退化重采样技术,将每个粒子的当前适应度值与其自身的个体最优值进行比较,如果优于个体最优值,得到粒子当前最优位置。仿真实验表明,采用该算法,收敛速度很快,在粒子群进化50代以内就可以实现成功收敛,对流量序列的测度特征提取结果准确,预测精度较高,展示了算法的优越性能。
The network concurrent flow characteristics with signal time predictability, based on network flow solution of con?volution feature extraction, improve the performance of the prediction model of network traffic. The traditional method using particle swarm optimization algorithm to achieve the characteristic measurement of network traffic analysis of blind decon?volution, the statistical information of the original signal extraction effect is not good. Put forward a kind of particle swarm degradation of network traffic measure resampling deconvolution algorithm based on the construction of network traffic, and hair sequence acquisition model, design of particle degradation resampling technique, the current best individual fitness value of its own to compare the value of each particle, if better than individual optimal value, get the current best particle. Simulation results show that, by using this algorithm, the convergence speed is very fast, in the particle swarm evolution within 50 generations can achieve successful convergence, measure characteristics of flow sequence extraction result is ac?curate, the prediction accuracy is higher, it has the superior performance of the algorithm.
出处
《科技通报》
北大核心
2015年第6期196-198,共3页
Bulletin of Science and Technology
基金
北京市教育委员会基金资助项(KM201000002003)
关键词
智能粒子群
采样
网络流量
测度
intelligent particle swarm
sampling
network flow
measure
