Technical Data Stream Flow Recognition Mining

With the rapid development of internet technology, all kinds of new typed networkapplication increased network traffic sharply, which brought a huge challenge for bandwidth’sdisorder immoderate preemption and the network bandwidth resources management. Networktraffic identification technology research solved the decrease of service quality due tounreasonable using of bandwidth resource, which provided a scientific grounding for real-timemonitoring the use of different business quality, link load balancing, reducing the waste ofbandwidth resources and increasing key business transmission efficiency. In addition, networktraffic identification also played an important role in sorts of network accounting, networksecurity and traffic engineering.This article analyzed the current research situation of traffic identification technologyfirstly. The evaluation criteria and recognition results’ evaluation indicators of network trafficidentification technology was given. The existing problems in the recent high-speed networktraffic identification was also analyzed. And then the paper also highlighted traffic identificationtechnology on the base of data stream mining. In the end, the paper designed and realized trafficidentification system on the base of data stream mining, and tested the system of campusnetwork. The main research content and innovative results of the paper are as follows.(1)This article applied CVFDT algorithm to the network traffic identification and analyzedhow to apply CVFDT algorithm to the network traffic identification in detail, and we use theauthoritative data through experiments comparing the classification accuracy CVFDT algorithmsand VFDT algorithm and the ability to solve concept drift.(2)The paper adopted CVFDT algorithm to design real-time traffic identification system,and given the overall design of the system and the method for implementing key technology offlow collection, sampling, flow convergence, flow real-time attribute selection, building updecision tree and class mark. The system adopted capture method libpcap packet-based and thesampling technique was applied to the packet acquisition process, because software acquisitionpackets mismatched the speed of the network packets resulted in the severe packet loss rate,sampling technique was to use simple random sampling method; Flow convergence determinedwhether to belong to a stream by comparing the packet five-element group, and then gathered thepacket into a stream. Flow real-time attribute selection chose the seven attributes with real-time,genetic diversity. Constructing a decision tree was according to CVFDT algorithm to build anupdate decision tree. Classification module took the seven attribute values of real-time attributeselection module as input and completed class mark on traffic. In the campus network, We testedthe traffic classification system by the collection and reorganization of the flows. It assisted thestatistic and prediction of the usage of network effectively.