Analysis And Research On The Spatio-Temporal Pattern Of Urban Crowd Activities Based On Mobile Signaling Data

With the rise of the global information technology wave and urban computing concepts,the potential value mining of emerging big data and data-driven scientific research strategies have increasingly received widespread attention from the academic community,and have been applied to the exploration and decision support of practical problems in human behavior,sociology,urban planning and other disciplines The rapid development of urbanization is the current trend of the times,especially in recent years,the urbanization process in developing countries is accelerating,leading to the spread of diseases,ecological damage,traffic congestion and other urban problems emerge in endlessly.Therefore,how to effectively use emerging perceptual analysis methods such as ubiquitous location information,big data mining,and deep learning,and give full play to the advantages of multidisciplinary integration of geographic information science,build an urban activity analysis model for a large number of people,and explore various urban activities The spatial-temporal distribution model has become a key technical means to realize the dynamic monitoring of complex urban social systems and the analysis of related issues.On the one hand,it can provide methodological support for the relevant analysis of social and humanities,and at the same time,it can also provide a scientific source of information for the development status and problem discovery of urban population,transportation,medical care,leisure and tourism industries.At present,some scholars at home and abroad have carried out research on urban crowd activities based on ubiquitous location data.However,due to the limitations of data sources and analysis techniques,most of these studies focus on the estimation of population or population density distribution.From the perspective of time scale,day is the most analytical scale;From the perspective of space scale,there are more studies on the unit of kilometer network or administrative division;From the perspective of activity type semantics,most of them focus on the analysis of traffic,family and other specific spatial scenes,but the modeling and analysis method of urban multi type activity distribution has not been proposed.Combined with the shortcomings of the existing research,in the face of the low value density of big data and the diversity and complexity of human behavior,this study comprehensively uses data mining technology,behavior analysis technology,probability graph learning technology and spatial analysis technology,takes the spatial-temporal pattern analysis of urban crowd activities as the research object,and the research framework of human activity modeling,type inference and distribution estimation model is systematically proposed,and the urban modeling and its application mode research centered on crowd activities have been realized,which mainly including the following three aspects :(1)This paper systematically studies the diversity of human activities in the city,and solves the problem of how to build a human city activity model based on massive mobile phone signaling data.Due to the differences of age,occupation and gender,people's urban activities are rich and colorful,and the classification of activity types is also different according to the research purpose and the granularity of modeling.This paper puts forward the main classification of urban human activities based on the questionnaire survey and the related research of human urban behavior.At the same time,facing the huge urban system with high population density,by analyzing the advantages and disadvantages of all kinds of big data in human activity modeling,the mobile phone location data with certain spatial-temporal accuracy,long-time observation and high population coverage is selected as the data source.Combined with the characteristics of mobile phone location data,comprehensively using methods such as Hadoop cluster framework and speed-based filtering to construct and implement a human city activity chain model.(2)Proposed and realized the semi-supervised city activity type recognition probability graph neural network model and the comprehensive multi-dimensional auxiliary urban spatiotemporal crowd activity distribution interpolation method.In view of the difficulty of converging the posterior distribution of complex relationship patterns in the current statistical relationship learning model in the inference of urban activity types,and the dependence of the relationship model's potential function on manual construction,the ability to express data features is very limited,a semi supervised graph Markov network model is proposed to solve the activity type annotation task with sparse samples and complex graph relationship.In addition,considering that most of the current human activity distribution estimates do not take into account the semantics of activities,this article makes full use of the advantages of multi-source big data,and uses multi-source data to construct multi-dimensional auxiliary information as effective space-time constraints and weight coefficients,and then constructs the estimation model of urban activity spatial-temporal distribution pattern.(3)Explore the activity pattern construction technology of multi-temporal and spatial scales to improve the robustness of the model and adapt it to the needs of diversified real-life application scenarios.At the same time,it proposes the application scope and application mode of the construction technology of urban population activity distribution pattern,and explores its application value in the research fields of urban geography,tourism geography and other disciplines.On the one hand,the inference method of urban population activity types is applied to the evaluation of urban vitality,to construct an assessment framework for the connotation of urban vitality,including the assessment of economic vitality and social vitality,and then to explore the driving factors and shaping strategies of urban vitality.On the other hand,the estimation model of the spatial-temporal distribution pattern of urban population activities is applied to the analysis of urban leisure patterns,to explore the activities of urban populations in the third space,and through the mining and comparative analysis of the activity space pattern and type preference in different periods,suggestions for the development of urban leisure space and future planning suggestions are put forward,aiming to promote the coordinated and sustainable development of the city.