User Behavior Analysis And Mobility Prediction Based On Mobile Communication Data

With the improvement of material living standards,tourism as an important part of people's spiritual life,the demand for tourism has also been rapidly improved.How to better serve tourists has become a key issue for governments and enterprises.Nowadays,the rapid development of the communication industry and the construction of facilities such as the LTE signaling monitoring system provide the tourism industry with a means to obtain user movement traj ectory data,and operators can more easily obtain interaction data between users and base stations.Identifying tourists and forecasting the itinerary of tourists is conducive to the development of tourism services and helps the government and enterprises to rationally allocate tourism resources.This paper will study the two maj or issues in the tourism industry,including tourist identification and tourist travel forecasts.In order to support the analysis of tourist identification and tourist behavior prediction,this paper first studies the related processing technology of mobile data,including data filtering,de-abnormal,data compression and other data pre-processing techniques;and attempts to optimize the processing flow,combining the drift effect and ping-pong effect.In the issue of tourist identification,based on the heuristic rule screening,a tourist identification scheme based on RNN neural network combined with heuristic rules was proposed,which improved the recognition rate of tourists by 1 1%.In the tourist travel forecasting problem,this paper enriches the scenic spot data through the network crawler,extracts the functional feature vector of the scenic spot,and compares the traditional machine learning classification algorithm,it is verified that the RNN neural network can be used in the single-step prediction problem of tourists' travel.In the multi-step prediction of tourists' journey,a method of trajectory sequence similarity comparison is proposed.The feature vector of the scenic spots'features is also used as a dimension to participate in the calculation of trajectory similarity.The absolute time and time duration of the event in the trajectory are simultaneously involved in calculation of trajectory similarity.At the same time,this similarity calculation method is applied to the multi-step prediction problem of tourists' journey,and the application of various RNN neural networks based on attention mechanism in multi-step travel prediction is explored.