Diversity and complexity of network traffic and the high real-time requirement has brought great challenges

meanwhile by the network traffic classification based on machine learning

the selection of flow statistics feature can quickly and accurately identify traffic

which is of great significance. As a gray box model

random forest（ RF） is popular in recent years

which has good recognition ability of the current network traffic. By the algorithm of RF node selection

flow characteristics are selected

and combined with the physical significance to carry out optimization and a number of feature extraction. Experimental results show that the proposed algorithm is superior to the model construction

time-consumption is greatly reduced

meanwhile the accuracy is improved.