| Nowadays,As mobile terminal devices are widely used in people’s life,The number of mobile Internet applications is growing rapidly.It is a great challenge for network managers to identify and classify these flows to facilitate management and control.Because mobile traffic has its own characteristics,which makes most of the identification and classification methods can be used in PC’s flow too limited to be used in traffic in the mobile Internet applications.In order to solve the above problems,this thesis studies the two directions of eigenvalue and algorithm design,and seeks the breakthrough.Firstly,this thesis analyzes the communication protocol of mobile Internet application,and uses Wireshark and other tools to grasp mobile Internet protocol traffic.By analyzing the content of flow and its form in the network transmission,combining with the basic research in the field,designing experiments from a variety of angles,observing the flows and exploring the abstract nature of flows’ attribution.Then the thesis put these abstract natures into a characteristic value as a feature of algorithm design on feature selection,creatively introduced and message header information entropy as a new characteristic value of flow and its measurement method.Next,this thesis designs a mobile Internet traffic identification and classification scheme based on machine learning and realizes the prototype system of the scheme.In the design of the algorithm,this thesis determines the scheme combination of classification and clustering.The semi-supervised learning S4 VM classification algorithm is proposed in this thesis effectively to reduce the cost of training sample collection.At the same time,the AGNES clustering algorithm proposed in this thesis can ensure that the scheme can achieve better traffic coverage on the basis of application coverage.This thesis also introduces the feedback and grouping mechanism to combine the two algorithms and improve their performance.Finally,designing experiment verifies the reliability and efficiency of the scheme. |
