Abstract:P2P software is widely used in networks．It is a very important issue that how to identify P2P data flows fast and effectively. The traditional P2P identification methods are not suitable to widely applied dynamic port and the content encryption. Combining with data mining classification, this paper proposes a new decision algorithm based on distance decision function according to the dynamic characteristics of P2P flows, including IP packets numbers，percent of UDP and link-numbers. The experiment results show that P2P data flows can be identified and predicated quickly and effectively using this algorithm, therefore P2P flows can be monitored.