Research And Application Of Urban Planning Modeling Based On Mobile Signaling Big Data

With the popularization of mobile phone devices and the continuous advancement of big data processing technologies in recent years,there are more and more applications based on mobile phone signaling big data.The urban planning has become a hot topic in big data research with the urban problems such as traffic congestion,environmental pollution,and unreasonable infrastructure planning.The characteristics of crowd flow are extremely significant for urban traffic management,public safety and basic transportation resource planning.Based on the mobile phone signaling big data,the thesis focuses on the research of urban crowd flow prediction and urban congestion modeling.The main work of the thesis is as follows:1.Data preprocessing process and space-time index structure are designed.According to the characteristics of mobile phone signaling data and the application in urban planning modeling,a set of data cleaning and preprocessing procedures for mobile phone signaling data are designed based on the research on spatio-temporal data preprocessing methods.The quality of the data is improved and the processing complexity is reduced by cleaning the invalid data,drift data and ping-pong switching data in the original data.Then,according to the spatio-temporal characteristics of the mobile phone signaling trajectory data,trajectory compression,trajectory stay point analysis,traffic mode analysis and trajectory segmentation are carried out for mobile signaling trajectory data.On the one hand,our works improve the quality of mobile signaling trajectory,on the other hand,they enrich the semantics of mobile signaling trajectory,which is conducive to further mining work.A spatio-temporal index structure based on spatial partitioning and B+ tree is also designed,which greatly improves the retrieval speed of data.2.An urban crowd flow prediction algorithm is proposed.Firstly,we provide the definition of urban area and crowd traffic.Based on the historical trajectory data of mobile phone signaling,the historical crowd flow data of the urban area per unit time is obtained.Then,the deep space-time residual network and local convolution are applied to set up the prediction Model framework of urban crowd flow called PS-ResNet,by simulating weather,events and the three characteristics that affect the future crowd flow time.The three characteristics include similarity,period and trend.Finally,We do a series of experiments on PS-ResNet.Compared with some population prediction algorithms,the result of PS-ResNet is proved to be more precise and stable.3.Established a prediction model for urban congestion.Firstly,based on the theory of space-time consumption,a model of urban road network traffic carrying capacity is established to quantitatively describe the comprehensive traffic capacity of road network.Then,based on the PS-ResNet algorithm framework,we predict the future traffic flow of the road network.Combining with the road network traffic carrying capacity,we get the short-term road network traffic saturation value in the future.Finally,according to the classification of service level of China's road network,combined with the prediction result of saturation,we put forward short-term and long-term rationalization proposals for optimizing traffic capacity to the urban planning and traffic management departments.