P2P Network Traffic Identification Based on Random Forest Algorithm

With the rapid development of computer technique in the past decades, the emergence of P2P techniqueprompts the network computing model evolving from centralized network to distributed network. Although P2P technique has brought tremendous changes to the network technique, P2P technique also exposes a lot of problems during its implementation. If we can manage the P2P network traffic effectively,e.g. identifies and controls its traffic and distinguishes its services, then it will make great sense for research on improving the performance of network service and use efficiency. However,the traditional approaches have shown a great lack of adaptability in dealing with samples which contain heterogeneous information.large scale of samples,unnormalizeddata or uneven data distributed in high dimensional feature space. This paper is based on therelated researches, to overcomethe limitations and shortcomings of current network traffic identification; we explored network traffic identification and came up with an approach of network traffic identification based on random forest. This paper uses campus network of North China University of Water Resources and Electric Power and takes its outlet flow as sample data to experiment. The result shows that random forest is suitable for large scale of data situation, complex dimensional situation, data contain lots of heterogeneous information etc. Additionally, random forest algorithm provide broad application prospects and rich design ideas for machine learning in feature extraction, multiple class object detection and pattern recognition fields.