Research On Identification Method Of Large-scale Network Traffic Based On Machine Learning

From the 1990 s to the present,the Internet has experienced a stage of rapid development,and great changes have taken place in peopleundefineds work and way of life.The demand for users to use the Internet is rising at an extreme speed.By detecting different network traffic,we can adopt corresponding management means according to the influence of traffic produced by different applications on the network,so as to realize targeted traffic management,network optimization,situation prediction and security prevention.It has important significance for reasonable planning of network resources,reduction of operation cost,maintenance of network security and improvement of quality of service.In this paper,the traffic recognition method based on machine learning is adopted,and the support vector machine is selected as the machine learning algorithm for traffic recognition.Since the kernel function is introduced into the support vector machine(SVM),the algorithm can transform the low-dimensional linear non-separable problem into the high-dimensional linear separable problem.The algorithm complexity is reduced and the high-dimensional disaster problem is avoided.SVM follows the principle of minimum structural risk,improves the generalization ability of classifier learning machine and avoids the problem of over-fitting in the training process.The flow recognition method based on support vector machine can avoid falling into the local optimal solution and obtain the global optimal solution without training large sample data.In order to solve the problem of high dimensional traffic feature disaster,this paper proposes a hybrid feature selection algorithm based on support vector machine(SVM),and compares it with the SVM traffic recognition algorithm under the traditional feature selection method.The experimental results show that the feature selection algorithm proposed in this paper has excellent recognition performance.Another important research topic based on machine learning traffic recognition is multi-category classification.When the number of application types increases gradually,the number of binary classifiers will also increase sharply,making the decision efficiency of SVM classifiers significantly lower.This paper proposes a multi-class classification model of SVM based on optimized error-correcting codes.In this paper,multi-group and multi-class classification experiments are carried out,and the recognition performance of the traditional multi-class classification model and the optimized multi-class classification model of error-correcting codes is compared.The experimental results show that theperformance of the multi-class classification model is always better than that of the traditional multi-class classification model with the same number of traffic classes.And with the increase of the number of traffic categories,the former identification performance tends to be stable,while the latter identification performance decreases.Therefore,in the traffic identification of this paper,the SVM multi-class classification model based on optimized error-correcting codes can provide stable and excellent recognition performance.