Abstract: In order to improve the identification effect of network communication load state, we proposed a identification model of network communication load state based on relevance vector machine. Firstly, the parameters which affected the quality of network communication were extracted, and the relationship between them and their load states was analyzed. Secondly, the throughput of wireless sensor network was regarded as the standard of load state identification, and the relevance vector machine was used to construct the classifier of network communication load state to realize the identification of network communication load state. Finally, the specific data was used to test the performance of network communication load state identification. The test results show that the relevance vector machine can accurately identify the network communication load state, and the correct rate of the network communication load state identification is higher than other models.

Key words: wireless sensor networks, network throughput, relevance vector machine, network communication load; identification model