The human are living in a sense network environment due to the development of a variety of sensing technologies. A number of digital information which contains lots of valuable human behavior patterns may be left behind anywhere and at anytime. How to take advantage of the information has been a major research field in information science with the purpose of observing inherent laws of human behavior, establishing the cognitive structure and serving the human society better. This paper explores new methods for behavior modeling and behavior analysis to mine the laws hidden in mobile data in the background of city transportation, thus not only has important theoretical value, but also has major operation significance. This paper presents a research framework of sense network from three aspects of contents, methods and application and points out the problems of the current research. The research framework here can give us a comprehensive understand of sense network. Influence model is the theory for several interactive stochastic processes. After explaining the basic theory of influence model, this paper constructs the influence model for singles and swarms in sense network which can be used to behavior modeling and behavior analysis, then summarize the general steps for modeling and the characteristics of problems which can be solved by this theory. Forecasting traffic flow in a high accuracy is the key to intelligent transportation. This paper describes the traffic flow forecasting problem, analyzes the characteristics of traffic flow with an emphasis on the qualitative and quantitative proving of influence between nodes in transportation network, and gives the requirements of the methods for traffic flow forecasting. According to the characteristic that traffic flow of different nodes influences each other in complex traffic network; this paper presents a novel short-term traffic flow forecasting method based on influence model. As the traffic flow of each node is considered as a hidden markov process, the whole traffic network is composed of many interactive hidden markov processes. This method uses the EM algorithm to learn parameters in the forecasting model. Experimental results show that the method proposed in this paper not only has higher prediction accuracy, but also presents the interactive influence and dynamic evolution of the traffic flow of different nodes in traffic network.