关键词: 流量分类, 统计特征, 机器学习, 自组织映射

Abstract: In order to satisfy the requirements of users for more and more precise Internet service quality, the traffic classification is an important link in the network management process. Through analyzing and comparing the application situation and the advantages and disadvantages of each classification method by machine learning, which were separately based on port number matching, feature analysis and traffic characteristics, a hybrid model of network traffic classification method was proposed to solve the problems that rely on a single classification method, such as low accuracy, long classification time. This model combined the port number matching with machine learning, and applied Self-Organizing Map (SOM) of which the output result is visual. The experimental result shows that this method can effectively achieve the application type classification of network traffic, and obtain a good visual effect of classification result.

Key words: traffic classification, statistical characteristic, machine learning, Self-Organizing Map (SOM)