A novel model for user clicks identification based on hidden semi-Markov

User click pattern is a critical part of cyber behavior, which is important for web operators and designers. Usually, massive mixed HTTP requests, we observed, are caused by the clicks from numerous web browsers. Moreover, the ever growing complexity of the web makes it more difficult to identify user click pattern accurately. In this paper, we propose a novel user click pattern identification method based on the hidden semi-Markov model. Moreover, we develop the parameter estimation and state estimation algorithms for our model. In order to initialize the model state value and improve the applicability of our method for real websites, we propose a state selection algorithm based on K-means clustering to reveal the in-line objects of web pages on different websites. We evaluate our method with a real data set, which is collected at the backbone of a state Telecom. The experiment results demonstrated that our method works quite well.