Using PCA algorithm to refine the results of internet traffic identification

Detecting and identification the network traffic attracts many attentions in recent years. Statistical approach using the machining learning algorithm can classify the network traffic efficiently without detecting the payload of every packet. At the same time, the accuracy depends on the statistical features of the training set. However, the traditional process without pre-treatment of the statistical features can lead to the misidentification in many scenarios. In this paper, an improved method is proposed based on the PCA (Principle Component Analysis) algorithm for pre-treatment of the statistical features, which is able to refine the results of traffic identification. Extensive experiments have been done and the results show that the accuracy rate of traffic classification based on the improved statistical method is improved.