网络流量预测系统的研究与实现
|
摘要
目前网络行为特征日趋复杂,随着Internet飞速发展,这给网络规划、网络管理带来了巨大的挑战,矛盾也日益突出,进行网络流量分析进而建立有效的网络流量模型是当前网络研究的热点问题之一。网络流量模型是对网络流量特性的刻画,一个好的模型不仅应能准确地反映出历史流量的特性,而且还能有效预测将来一段时间内网络流量的大小。一些传统模型通常假定网络流量满足线性关系,然后用线性递推和组合的方法来描述系统。实际上网络流量数据随时间变化并未表现出显著规律,因为它包含了很多非线性因素。本文围绕提高预测精度和自适应能力的主题,在网络流量预测中采用非线性模型,作了深入的分析,并在此基础上设计并实现了网络流量预测分析系统。
本文在引入时间序列预测理论和人工神经网络理论的基础上,设计出基于BP神经网络的流量预测模型,给出了有效流量预测方案。本文在预测模型中加入了动态调用预测功能,使得所设计网络流量预测系统在灵活性和高效性方面得到了提升,并且具备较强的自适应能力。
本文所设计的预测系统通过大量实验获得最佳预测模型,应用该系统能有效预测短期内网络流量的变化,进行信息预报,为网络规划、网络管理做出了积极有益探索。
At present, the features of network behavior are becoming increasing complicated, which, with the rapid development of Internet, brings a great challenge to the network planning and network management. One of the present hot points is to create an effective network flow model for the network flow analysis. The network flow model is a depiction of the features of the network flow. A good model not only should accurately reflect the features of historic flow but also can predict the size of the network flow in a time to come. Some traditional models usually assume that the network flow meets with the linear relationship and then describe the system with the linear recursion and combination. In fact, the network flow data don’t show the remarkable law with the change of the time, because they contain many non-linear factors. This paper, around the self-adapting ability and increasing the accuracy of prediction, makes a detailed analysis with the non-linear model in the network flow prediction, and designs and realizes an analyzing system of network flow prediction on that basis.
The network flow prediction system introduced in this paper, on the basis of introducing the theory of time series prediction and the theory of artificial neural network, designed a flow predicting model based on the BP-neural network, offering an effective predicting solution. The dynamic call predicting function in the predicting model improves the flexibility and efficiency of the system introduced in this paper, which is also of a stronger self-adapting ability.
This system acquires a best predicting model through much experiment; by means of this system, an effective prediction can be made on the network flow in the short-term period and on the information, which offers an active and beneficial exploration in the network planning and network management.
本文在引入时间序列预测理论和人工神经网络理论的基础上,设计出基于BP神经网络的流量预测模型,给出了有效流量预测方案。本文在预测模型中加入了动态调用预测功能,使得所设计网络流量预测系统在灵活性和高效性方面得到了提升,并且具备较强的自适应能力。
本文所设计的预测系统通过大量实验获得最佳预测模型,应用该系统能有效预测短期内网络流量的变化,进行信息预报,为网络规划、网络管理做出了积极有益探索。
At present, the features of network behavior are becoming increasing complicated, which, with the rapid development of Internet, brings a great challenge to the network planning and network management. One of the present hot points is to create an effective network flow model for the network flow analysis. The network flow model is a depiction of the features of the network flow. A good model not only should accurately reflect the features of historic flow but also can predict the size of the network flow in a time to come. Some traditional models usually assume that the network flow meets with the linear relationship and then describe the system with the linear recursion and combination. In fact, the network flow data don’t show the remarkable law with the change of the time, because they contain many non-linear factors. This paper, around the self-adapting ability and increasing the accuracy of prediction, makes a detailed analysis with the non-linear model in the network flow prediction, and designs and realizes an analyzing system of network flow prediction on that basis.
The network flow prediction system introduced in this paper, on the basis of introducing the theory of time series prediction and the theory of artificial neural network, designed a flow predicting model based on the BP-neural network, offering an effective predicting solution. The dynamic call predicting function in the predicting model improves the flexibility and efficiency of the system introduced in this paper, which is also of a stronger self-adapting ability.
This system acquires a best predicting model through much experiment; by means of this system, an effective prediction can be made on the network flow in the short-term period and on the information, which offers an active and beneficial exploration in the network planning and network management.
引文
[1] Abdelnaser Adas. Traffic models in broadband networks. IEEE Communications Magazine, 1997, 35(7), pp. 82-89.
[2] B. S. Chen, S. C. Peng and K. C. Wang. Traffic Modeling, Prediction, and Congestion Contol for High-SpeedNetworks: A Fuzzy AR Approach. IEEE Trans. Fuzzy Systems, Oct. 2000 Vol.8 No.5 P.491-508.
[3] V. Paxson and S. Floyd. Wide area traffic: The failure of Poisson modeling. IEEE/ACM Trans. Networking, June1995, vol. 3, pp. 226-244.
[4] N. Akar and E. Arikan. Markov modulated periodic arrivalprocess offered to an ATM multiplexer. Perform. Eval.,1994, vol. 22, pp. 175-190.
[5] V. N. Bhat. Renewal approximations of the switched Poisson processes and their applications to queueingsystem. J. Operational Res. Soc., 1994, vol. 45, pp.345-353.
[6] 邹柏贤,刘强. 基于 ARMA 模型的网络流量预测. 计算机研究与发展, 2002.12, Vol.39 No.12 , pp. 1645-1652.
[7] KHOTANZAD A, SADEK N. Multi-scale high-speed network traffic prediction using combination of neural networks[J]. Proceedings of the International Joint Conference on Neural Networks, 2003, (2):1071-1075.
[8] SADEK N, KHOTANZAD A. Dynamic Bandwidth Allocation Using A Two-Stage Fuzzy Neural Network Based Traffic Predictor[A]. International Joint Conference on Neural Networks, IJCNN’04[C],Budapest, Hungary, 2004.
[9] CHEN BS, PENG SC, WANG KC. Traffic modeling, prediction and congestion control for high-speed networks: A Fuzzy AR Approch[J]. IEEE Trans. On Fuzzy System, 2000, (8): 491-508.
[10] 郝昱文. 网络流量分析系统的设计与实现. 北京邮电大学, 2005.2.
[11] Paxson V, Floyd S. Wide-area Traffic: The failure of Poisson modeling [J]. IEEE/ACM Trans, Networking, 1995, Vol. 3, pp. 226-244.
[12] Lapedes A, Farber R. Nonliner signal processing using neural networks. Predictionand System Modeling, Tech LA-UR-87-2662, Los Alamos, Los Alamos National Laboratory, 1987.
[13] Groschwitz N.K., Polyzos G.C.. A time series model of long-term NSFNET backbone traffic. In: Proceedings of IEEE ICC, Pittsburgh, PA, 1994, 234-238.
[14] Sang A. ,Li S.. Predictability analysis of network traffic. In: Proceedings of INFOCOM, TelAviv, Israel, 2000, 342-351.
[15] FENG HA, SHU YT. Study on Network Traffic Prediction Techniques [A]. Wireless Communications, Networking and Mobile Computing Proceedings[C], 2005, 2:995-998.
[16] C.Quek, A.Singh.POP-Yager. A novel self-organizing fuzzy neural network based on the Yager iferece. Expert Systems with Application, 2005, 29:229242.
[17] 洪飞, 吴智美. 基于小波的多尺度网络流量预测模型. 计算机学报, 2006.
[18] 王新. 自相似网络流量的建模与预测[D]. 北京:清华大学电子工程系, 2003.
[19] 丛锁, 韩良秀. 基于离散小波变换的网络流量多重分形模型 [J]. 通信学报,2003.8, Vol.24 No.5 P.3-8.
[20] 邹柏贤, 姚志强. 一种网络流量平稳化方法[J]. 通信学报, 2004.8, Vol.25 No.8 P.14-23.
[21] 冯海亮, 陈涤等. 一种基于神经网络的网络流量组合预测模型. 计算机应用, 2006.
[22] 贾俊平, 何晓群等. 统计学[M]. 中国人民大学出版社, 2000.
[23] 向小东. 基于神经网络与混沌理论的非线性时间序列预测研究[D]. 西南交通大学, 2001.12.
[24] 张玉瑞, 陈建波. 基于 RBF 神经网络的时间序列预测[J]. 计算机工程与应用, 2005,Vo1.41,No.11:74-76.
[25] Lapedes A, Farber R. Nonlinear Signal Processing Using Neural Networks. Prediction and System Modelling[J]. IEEE Transactions on Neural Networks, 1995, Vol.4,No.2:169-171.
[26] Kolmogorov A. N. On the Representation of Continuous Function of Many Variables by Superposition of Continuous Functions of One Variable and Addition[J]. American Mathematical Society Translation, 1963, 28:55-59.
[27] Hecht Nielsen R. Neurocomputing[M]. Addison Wesley, 1990.
[28] 肖雪峰, 王建新, 陈建二.网络流量测量中相关参数的自回归预测分析. 计算机工程, 2006 Vol.32 No.6 P.130-133.
[29] 陆国浩, 朱艳琴. 人工神经网络应用于时间序列预测的探讨. 沙洲职业工学院学报, 2006 Vol.9 No.3.
[30] MRTG. http://oss.oetiker.ch/mrtg.
[31] RRDTOOL. http://oss.oetiker.ch/rrdtool.
[32] 刘芹, 余一娇, 谭连生. 一种利用 BP 神经网络的 Internet 流量预测算法. 中国计算机大会, 2003.
[33] 杨凡, 赵建民等. 一种基于 BP 神经网络的车牌字符分类识别方法[J]. 计算机科学, 2005,Vol.32,No.8:192-195.
[34] 张立明. 人工神经网络的模型及其应用[M]. 复旦大学出版社, 1993.
[35] Varfis A, Versino C. Univariate Economic Time Series Forecasting by Connectionist Methods[J]. IEEE, CNN-90, 1990,342-345.
[36] 高隽. 人工神经网络原理及仿真实例[M]. 机械工业出版社, 2003.8.
[37] Nguyen D., Widrow B. Improving the learning speed of 2-layer neural networks by choosing initial values of the adaptive weights[J]. Proceeding of the International Joint Conference on Neural Networks, Washington, 1990,Vol3:21-26.
[38] 董长虹. Matlab 神经网络与应用[M]. 国防工业出版社, 2005.1.
[39] 秦绪伟, 董傲霜, 洪智勇. 基于改进 BP 算法的商品销售量的预测. 沈阳工业学院学报, 2004 Vol.23 No.4 P.83-86.
[40] DE, Rumelhart, GE. Hinton Learning representations by back-propagation errors[J]. Nature, 1986, 323:533-536.
[41] RA Jacobs. Increased rates of convergence through learning rate adaptation[J]. Neural Networks, 1988, vol.1,No.4:295-308.
[42] Dilip Sarkar. Methods to speed up error BP Learning Algorithm[J]. ACM computing survey. 1995, Vol.27,No.4:519-592.
[43] Leonard J etc. Improvement of the Back-propagation Algorithm for Training Neural Networks[J]. Computers Chem. Engng, 1990, Vol.14,No.13: 337-341.
[2] B. S. Chen, S. C. Peng and K. C. Wang. Traffic Modeling, Prediction, and Congestion Contol for High-SpeedNetworks: A Fuzzy AR Approach. IEEE Trans. Fuzzy Systems, Oct. 2000 Vol.8 No.5 P.491-508.
[3] V. Paxson and S. Floyd. Wide area traffic: The failure of Poisson modeling. IEEE/ACM Trans. Networking, June1995, vol. 3, pp. 226-244.
[4] N. Akar and E. Arikan. Markov modulated periodic arrivalprocess offered to an ATM multiplexer. Perform. Eval.,1994, vol. 22, pp. 175-190.
[5] V. N. Bhat. Renewal approximations of the switched Poisson processes and their applications to queueingsystem. J. Operational Res. Soc., 1994, vol. 45, pp.345-353.
[6] 邹柏贤,刘强. 基于 ARMA 模型的网络流量预测. 计算机研究与发展, 2002.12, Vol.39 No.12 , pp. 1645-1652.
[7] KHOTANZAD A, SADEK N. Multi-scale high-speed network traffic prediction using combination of neural networks[J]. Proceedings of the International Joint Conference on Neural Networks, 2003, (2):1071-1075.
[8] SADEK N, KHOTANZAD A. Dynamic Bandwidth Allocation Using A Two-Stage Fuzzy Neural Network Based Traffic Predictor[A]. International Joint Conference on Neural Networks, IJCNN’04[C],Budapest, Hungary, 2004.
[9] CHEN BS, PENG SC, WANG KC. Traffic modeling, prediction and congestion control for high-speed networks: A Fuzzy AR Approch[J]. IEEE Trans. On Fuzzy System, 2000, (8): 491-508.
[10] 郝昱文. 网络流量分析系统的设计与实现. 北京邮电大学, 2005.2.
[11] Paxson V, Floyd S. Wide-area Traffic: The failure of Poisson modeling [J]. IEEE/ACM Trans, Networking, 1995, Vol. 3, pp. 226-244.
[12] Lapedes A, Farber R. Nonliner signal processing using neural networks. Predictionand System Modeling, Tech LA-UR-87-2662, Los Alamos, Los Alamos National Laboratory, 1987.
[13] Groschwitz N.K., Polyzos G.C.. A time series model of long-term NSFNET backbone traffic. In: Proceedings of IEEE ICC, Pittsburgh, PA, 1994, 234-238.
[14] Sang A. ,Li S.. Predictability analysis of network traffic. In: Proceedings of INFOCOM, TelAviv, Israel, 2000, 342-351.
[15] FENG HA, SHU YT. Study on Network Traffic Prediction Techniques [A]. Wireless Communications, Networking and Mobile Computing Proceedings[C], 2005, 2:995-998.
[16] C.Quek, A.Singh.POP-Yager. A novel self-organizing fuzzy neural network based on the Yager iferece. Expert Systems with Application, 2005, 29:229242.
[17] 洪飞, 吴智美. 基于小波的多尺度网络流量预测模型. 计算机学报, 2006.
[18] 王新. 自相似网络流量的建模与预测[D]. 北京:清华大学电子工程系, 2003.
[19] 丛锁, 韩良秀. 基于离散小波变换的网络流量多重分形模型 [J]. 通信学报,2003.8, Vol.24 No.5 P.3-8.
[20] 邹柏贤, 姚志强. 一种网络流量平稳化方法[J]. 通信学报, 2004.8, Vol.25 No.8 P.14-23.
[21] 冯海亮, 陈涤等. 一种基于神经网络的网络流量组合预测模型. 计算机应用, 2006.
[22] 贾俊平, 何晓群等. 统计学[M]. 中国人民大学出版社, 2000.
[23] 向小东. 基于神经网络与混沌理论的非线性时间序列预测研究[D]. 西南交通大学, 2001.12.
[24] 张玉瑞, 陈建波. 基于 RBF 神经网络的时间序列预测[J]. 计算机工程与应用, 2005,Vo1.41,No.11:74-76.
[25] Lapedes A, Farber R. Nonlinear Signal Processing Using Neural Networks. Prediction and System Modelling[J]. IEEE Transactions on Neural Networks, 1995, Vol.4,No.2:169-171.
[26] Kolmogorov A. N. On the Representation of Continuous Function of Many Variables by Superposition of Continuous Functions of One Variable and Addition[J]. American Mathematical Society Translation, 1963, 28:55-59.
[27] Hecht Nielsen R. Neurocomputing[M]. Addison Wesley, 1990.
[28] 肖雪峰, 王建新, 陈建二.网络流量测量中相关参数的自回归预测分析. 计算机工程, 2006 Vol.32 No.6 P.130-133.
[29] 陆国浩, 朱艳琴. 人工神经网络应用于时间序列预测的探讨. 沙洲职业工学院学报, 2006 Vol.9 No.3.
[30] MRTG. http://oss.oetiker.ch/mrtg.
[31] RRDTOOL. http://oss.oetiker.ch/rrdtool.
[32] 刘芹, 余一娇, 谭连生. 一种利用 BP 神经网络的 Internet 流量预测算法. 中国计算机大会, 2003.
[33] 杨凡, 赵建民等. 一种基于 BP 神经网络的车牌字符分类识别方法[J]. 计算机科学, 2005,Vol.32,No.8:192-195.
[34] 张立明. 人工神经网络的模型及其应用[M]. 复旦大学出版社, 1993.
[35] Varfis A, Versino C. Univariate Economic Time Series Forecasting by Connectionist Methods[J]. IEEE, CNN-90, 1990,342-345.
[36] 高隽. 人工神经网络原理及仿真实例[M]. 机械工业出版社, 2003.8.
[37] Nguyen D., Widrow B. Improving the learning speed of 2-layer neural networks by choosing initial values of the adaptive weights[J]. Proceeding of the International Joint Conference on Neural Networks, Washington, 1990,Vol3:21-26.
[38] 董长虹. Matlab 神经网络与应用[M]. 国防工业出版社, 2005.1.
[39] 秦绪伟, 董傲霜, 洪智勇. 基于改进 BP 算法的商品销售量的预测. 沈阳工业学院学报, 2004 Vol.23 No.4 P.83-86.
[40] DE, Rumelhart, GE. Hinton Learning representations by back-propagation errors[J]. Nature, 1986, 323:533-536.
[41] RA Jacobs. Increased rates of convergence through learning rate adaptation[J]. Neural Networks, 1988, vol.1,No.4:295-308.
[42] Dilip Sarkar. Methods to speed up error BP Learning Algorithm[J]. ACM computing survey. 1995, Vol.27,No.4:519-592.
[43] Leonard J etc. Improvement of the Back-propagation Algorithm for Training Neural Networks[J]. Computers Chem. Engng, 1990, Vol.14,No.13: 337-341.