Research On User Mobility Model In Opportunistic Networks

In recent years,intelligent devices such as smart phones,tablets and other portable,low-cost,computing power and short distance wireless communication capabilities have been popularized rapidly,which has promoted the development of the mobile opportunistic network.The difference between the opportunistic network and the traditional network is that the mobile opportunistic network realizes the communication between the user nodes,which meet each other during the mobile process.Therefore,the most basic premise of the communication among the users in the opportunistic network is the need for the users to contact each other.It is important for researching content of mobile opportunistic networks by analyzing mobile characteristics and user mobility models in mobile opportunistic network.The proposed mobility model should match the actual mobile scenario to a maximum extent,so the most accurate and reliable mobility method is to collect mobile data from the mobile users in the real scene,and then analyze the mobility model.However,the current user mobile data acquisition and characteristics analysis system has a single network interface,low data acquisition dimension,and the collected data can not reflect the behavior of mobile users in an all-round way.This paper gives a design of a mobile data acquisition system and a user mobility model analysis system,in which the data acquisition system is implemented by the Java language,and its core modules are divided into Bluetooth scanning module,GPS model,Wi Fi scan module and cloud server module.At the same time,we use the collected data to analyze the mobile characteristics of users.The existing opportunistic network user mobility model can not predict the meeting of users well.Based on the real mobile data of mobile users,the paper proposes a scheme to predict the meeting behavior of mobile users by using the long short-term memory network.First,the mobile users' Bluetooth scanning data,GPS trajectory data and Wi Fi scanning data are collected.After processing,the meeting time sequence of mobile users is obtained,and the time series is quantized by using the Word2 Vec model.Then,the model of the long short-term memory network is built,and the user model is trained by user data.By testing and verifing data of 6 mobile users,the average prediction accuracy is 93.6%.The prediction results show that the prediction model of meeting mobile user based on long short-term memory network has high accuracy.On the basis of the probability prediction of the meeting of mobile users,in order to better describe the time and space information of the meeting of the mobile users,this paper proposes a prediction method of the meeting duration and the meeting place of the mobile users based on the hidden Markov model.First,on the basis of getting the meeting time sequence of the mobile users,by analyzing and processing the data of the mobile users,the sequence of meeting duration and the time series of meeting place corresponding to the meeting sequence of the mobile users are obtained.Then,the meeting sequence is observed as the meeting duration and the meeting place,respectively.The hidden Markov model is set up by the hidden state,and the model is also trained by the data,thus the corresponding meeting duration prediction model and the meeting location prediction model are obtained.Finally,the data of 6 mobile users collected are tested and verified.The meeting duration prediction accuracy of the mobile users is 72.1%,and the average location prediction accuracy is 80%.The prediction results show that the hidden Markov model both in meeting duration and location of model users is feasible.Through the above work based on the mobile user data in the real scene,this paper establishes a prediction model of the meeting between users.It realizes the prediction of the meeting duration and the meeting place of the mobile users by the prediction of the meeting probability in the opportunistic network,and.Due to the limited time and manpower,more time and more user nodes will be considered in the data collection in the future so as to better respond to the user mobility characteristics.At the same time,the algorithm model will be optimized to achieve better prediction accuracy.