Analysis Of Urban Travel Mode And Its Conduction Effect Based On Spatiotemporal Modeling

How to use city-aware big data to provide scientific and reasonable countermeasures and suggestions for many practical problems(such as traffic congestion,urban planning,etc.)in the development process has been a research hotspot in the field of urban computing in recent years.However,on the one hand,the spatial and temporal distribution of urban perception data is uneven,and the problem of "variable resolution" presents challenges to data repcesentation and characterization;on the other hand,urban multi-source perception data are independent of each other in their respective operating systems,and are related to each other in terms of semantic knowledge,so fully and effectively integrating multi-source perception data isn abother research challenge facing the urban computing field.This article focuses on urban transportation applications and comprehensively utilizes multi-source city perception data such as taxi trajectory data,Didi travel data,urban road network data,weather data,and points of interest,to analyze the travel time and space characteristics and laws of the citizen group in the urban environment,excavate the travel mode of the crowd under the multi-spatial-temporal granularity,analyze the traffic conduction effect,and apply the urban travel mode and its conduction effect to the prediction of traffic conditions.The research work of this paper mainly includes the following three aspects:1.Spatial and temporal multi granularity structured representation of city perception data.In the data preprocessing stage,aiming at the unbalanced spatial-temporal distribution of urban perceptual data,this paper proposes two kinds of spatial-temporal multi granularity structured representation methods,3dtree and subdivision merge,which avoid the low efficiency,semantic inaccuracy and low precision of traditional spatial-temporal data representation methods.At the same time,the proposed spatial-temporal multi granularity sructured representation method is verified by data of different types,different times and different regions The stability,universality and generalization ability of the data representation method with degree structure.2.Group mobility pattern mining and its transmission effect analysis in urban environment.Based on the spatiotemporal and multi granularity structure,combined with sequential pattern mining algorithm and frequent subgraph mining algorithm,a two-level pattern mining framework of region road network is proposed to mine the travel mode of urban residents and analyze the traffic conduction effect.Compared with the traditional methods,the patterns mined in this paper are rich and diverse,multi-source and heterogeneous,and have more attributes.3.The application of travel mode and its conduction effect in traffic condition prediction.Firstly,the embedding model is used to embed the spatiotemporal and multi granularity structure's spatiotemporal characteristics,traffic speed laws,crowd movement patterns and traffic conduction patterns into vector space.In the embedding process,the cbow model based on negative sampling is improved to make the embedding vector more accurate and the training efficiency higher.Then,the measurement learning is embedded into vector space Among the indicators of classifier accuracy,the accuracy of the method in this paper is improved by optimizing the indicators.