基于支持向量机的网络流量预测和资源调度

发布时间：2018-06-23 06:15

本文选题：支持向量机 + 核函数　；参考：《广东工业大学》2015年硕士论文

【摘要】：随着计算机和互联网的持续高速发展,网络在人们生活中扮演的角色也越来越重要,人们再也不能满足于只简单上网的需求,人们对上网的要求也越来越高。网络拥塞、网络故障、网络安全等一系列的问题时刻困扰着我们,如何对系统中的网络数据进行测量、收集和预测已成为网络系统运行的主要难题之一。据大量数据显示,网络是复杂的、多方因素影响的,网络流量也必然呈现出高度自相似、时变性和非线性等特征,这注定传统的预测方法无法做到高的准确率。支持向量机是一种机器学习方法,其求解速度快,且泛化能力强,故本文用支持向量机来进行预测。支持向量机可以根据现有的有限的样本信息,在所建立的模型的复杂性和机器的学习能力间寻求一个平衡点,以得到最好的泛化能力,并创造性的将线性不可分的问题,通过核函数映射到高维空间,使之线性可分。本文在对网络流量准确预测后,综合预测了CPU使用率和内存使用率的情况,为市区信访件对接平台设计了模糊控制器,该模糊控制器根据预测结果进行资源调度,并在仿真平台上进行了实验,取得了很好的效果。本文的主要研究内容如下：1).研究支持向量机参数选择的问题。参数的选择在支持向量机建模期间有巨大的影响,参数的好坏直接影响着预测精度的高低。在研究生学习期间,本人关注了各种新型的算法,并创新性的将布谷鸟搜索算法应用于支持向量机的参数选择过程中。实验对比了现有的算法,如遗传算法和粒子群算法,布谷鸟搜索算法明显提高了SVM的效率和结果准确率。2).根据记录的网络带宽、CPU使用率,内存使用率的数据,通过本文提出的基于布谷鸟搜索算法的支持向量回归机(CS-SVR)进行预测,并通过本文设计的模糊控制器根据CS-SVR的预测结果,对资源进行调度,使得服务器端的各项资源的利用率最大化,达到负载平衡,从而提高服务质量。
[Abstract]:With the continuous rapid development of computers and the Internet, the role of the network in people's lives is becoming more and more important. People can no longer meet the need of simply accessing the Internet, and people's requirements for the Internet are also getting higher and higher. A series of problems, such as network congestion, network failure, network security and so on, haunt us all the time. How to measure, collect and predict the network data in the system has become one of the main problems in the operation of the network system. According to a large number of data, the network is complex and influenced by many factors, and the network traffic must be highly self-similar, time-varying and nonlinear, which is doomed to the traditional prediction method can not achieve high accuracy. Support vector machine (SVM) is a kind of machine learning method, which has fast solving speed and strong generalization ability, so this paper uses support vector machine to predict. Support vector machine (SVM) can find a balance between the complexity of the established model and the learning ability of the machine based on the existing limited sample information in order to obtain the best generalization ability and creatively solve the problem of linear inseparability. The kernel function is mapped to high dimensional space to make it linearly separable. After the accurate prediction of network traffic, the CPU utilization rate and memory utilization rate are forecasted synthetically, and a fuzzy controller is designed for the docking platform of letters and visits in the urban area. The fuzzy controller schedules the resources according to the forecast results. Experiments are carried out on the simulation platform, and good results are obtained. The main contents of this paper are as follows: 1). The parameter selection of support vector machine (SVM) is studied. The selection of parameters has a great influence on the modeling of support vector machines, and the quality of parameters directly affects the accuracy of prediction. During the post-graduate study, I pay attention to various new algorithms, and creatively apply the cuckoo search algorithm to the parameter selection process of support vector machine. Compared with the existing algorithms, such as genetic algorithm and particle swarm optimization algorithm, the cuckoo search algorithm improves the efficiency and accuracy of SVM significantly. According to the recorded data of CPU utilization and memory utilization, this paper proposes a support vector regression machine (CS-SVR) based on cuckoo search algorithm, and uses the fuzzy controller designed in this paper to predict the CS-SVR. The resources are scheduled to maximize the utilization of each resource on the server side to achieve load balance and improve the quality of service.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP18;TP393.06

【相似文献】