Mobile Signaling Based Regional Traffic Flow Prediction In Urban Areas

In traffic management and public safety applications,the prediction of traffic flow has become an increasingly important research problem.On the one hand,map navigation APP can provide users with a smooth route through real-time traffic prediction,thus saving the user travel time and improving the user experience.On the other hand,the change in urban regional traffic flow obtained from the historical traffic information is of great significance for urban planning,traffic management and public safety.At present,the commonly data for traffic flow prediction is GPS data,although accurate,it is difficult to collect enough GPS data except from volunteers or floating car acquisition system.Compared with the GPS data,the mobile phone signaling data is available at the base station in cellular communication system.Though its data quality is not so good,its large amount of data,wide coverage and simple acquisition makes it a promising alternative.Obviously,regional traffic extraction and prediction technology on the basis of mobile signaling data is important in terms of both research exploration and practical application.Based on the analysis of the mobile phone location related signaling data generation,it is proposed to eliminate the error data(ping-pong switch,drift,repetition value)in the mobile phone location based signaling data.In order to characterize the regional traffic,this paper proposes a grid density based method to extract and assess the regional traffic data.It is shown that,grid processing and grid traffic extraction can be utilized to obtain the grid based traffic status.Then according to the characteristics of the grid based traffic density,the region growth algorithm can be derived.Finally,the derived region traffic results can be used as the input to the subsequent traffic prediction model.Secondly,by analyzing the timing characteristics of mobile signaling data and combining the regional traffic extraction results,as well as synthetically considering the advantages of neural network in time series prediction,in this thesis the LSTM neural network is proposed to devise a regional traffic prediction model.In addition to the input layer,hidden layer and output layer,the traffic prediction effect is improved by adding a new layer of network structure to take into account the spatial characteristics of the urban area in traffic flow prediction.Finally,the paper uses the real mobile phone signaling data for experimental verification,and the proposed prediction model is compared with the traditional time series prediction model.The experimental analysis results show that,compared with the traditional traffic prediction model,the prediction accuracy in urban area can be improved by using the proposed LSTM cyclic neural network based regional traffic prediction model.Hopefully,the related analysis work in this thesis will provide useful reference for further research on efficient area traffic prediction technology from mobile phone signaling big data.