Mobile Signaling Data Based Urbanresident Movement Prediction Andanalysis In Grid Scale

With the development of mobile communication technology,traffic information acquisition technology based on cell phone signaling data has attracted more and more attention,which is expected to be widely used in practical applications.In urban traffic management and planning applications,it is of great significance to explore the temperal-spatial distribution characteristics of residents,to obtain the occupation and residence location distribution,to extract regional human flow and to monitor urban traffic from the mobile signaling data.At present,GPS data is widely used in urban traffic analysis.The GPS based data is accurate,however,its availability is very limited.Although the positioning accuracy is poor,there are still many advantages of the mobile signaling data,for instance,the sample size is large,the real-time performance is high and the acquisition cost is low.Moreover,mobile signaling data can provide us the highly desired wide coverage in both temporal and spatial scale,which can effectively satisfy the analysis requirements of the whole city.Therefore,the mobile signaling data based urban spatial characteristics analysis and the human flow at different times and different directions are of important research values.Firstly,we analyze the positioning principle of mobile communication to highlight the characteristics,the pros and cons of mobile signaling data.In order to effectively remove those errorenous data and incomplete data,the data pre-processed is performed at first.Moreover,in order to effectively suppress the positioning inaccuracy and to weaken the uneven sampling interval of the positioning data extracted from the available mobile signaling data,a grid-forming pre-processing technique is proposed such that every specific positioning data will be generalized to its grid.On the basis of grid processing,the paper presents a grid-scale regional human traffic extraction algorithm,in which the traffic of each grid can be derived,so that the traffic of urban areas at different times and in different directions can be derived according for the sake of urban spatial characteristics analysis and human flow forecasting.Secondly,the thesis analyzes the urban spatial characteristics of residents from three aspects.First,based on the spatial and temporal distribution of urban population,the population distribution of the entire city is analyzed.Secondly,a multi-day long-term residence identification method based on POI statistics is proposed.The distribution of occupation and residence location in the whole city and its impact on urban traffic flow are analyzed.The rationality of the identification method is verified through data sampleexpansion.Thirdly,on the basis of the above analysis,the formation mechanism and evolution analysis of traffic congestion in urban densely populated areas are carried out.Based on human flow density,the grid-level regional traffic state classficiation is performed to obtain the regional traffic congestion status and the traffic state of the whole urban area,and to derive the traffic congestion evolution from the temperoal and spatial perspective.Finally,the causes of congestion are summarized and the solutions to congestion are briefly presented.Finally,based on the extracted regional traffic status information,the paper analyzes the predictability of regional traffic sequence.Considering the advantages of LSTM neural network in time series prediction application,the paper uses LSTM neural network to establish regional human flow prediction model.By combining the influencing factors and spatial types of regional traffic sequences,the paper adds these two external features to the model for training,and improves the prediction accuracy of the regional resident flow prediction model.The experimental results confirm us that,the external features have a certain impact on the prediction model,which also prove its reasonability,the feasibility and effectiveness of LSTM based prediction system.