Several Key Technologies And Applications Research On Urban Transportation Data Mining Based On Multi-layer Road Index

In recent years,as the human society has gradually entered the era of intelligent information,the mobile real-time LBS(Location-Based Service)are becoming an indispensable part of people's life,including the e-hailing systems,shared bicycle systems,logistics/takeaway,mobile payment systems,video recognition systems,high-precision maps and autonomous driving.At the same time,the provision of intelligent information services have also promoted the development of large-scale data processing and analysis technologies,such as customer portrait technology based on massive spatial and temporal data,precision advertising technology,and short-term traffic/route prediction technology.In the field of urban traffic system analysis,various types of data sets based on large-scale data sets(such as urban traffic network data,cell phone signaling data,vehicle GPS trajectory data etc.)are carried out."Data Driven" research is entering the mainstream..Compared with traditional analysis methods,data-driven research has the advantages of more comprehensive sample collection,deeper research content,and more convincing research conclusions,but there are also several technical difficulties and pain points,such as large-scale city map data error correction coding,Semantic map generation,large-scale GPS data error correction,efficient map matching,and reliable algorithm design for specific topics.This paper establishes a set of large-scale urban traffic data processing and analysis techniques and research paradigms based on the evolution of multi-level road index definitions,and combines several research application scenarios to verify the practicability and feasibility of this set of technologies.The specific research content is as follows:(1)Propose the concept of Multi-layer Road Index(MRI)and the algorithm of building a MRI system for refined map based on the Extended Intersection Continuity Negotiation(EICN)This part firstly analyzes the technical requirements of modern large-scale urban traffic data mining and system analysis,and then describes the solutions to several typical map data quality problems,and gives the multi-level road indexing system(MRI system)definition.Finally,combined with the traditional dual topology technical specifications,the refined map data continuity rules(3C criterion,Cross Continuity Conditions)and EICN algorithm were designed.The algorithm is verified by Xi'an map data.The MRI system established by the algorithm can solve the problem that the original map system cannot accurately provide the semantic information of the road network and serve for further system analysis and research.At the same time,the EICN algorithm makes up for the defect that the traditional algorithm cannot handle the highly complex and refined road network information.(2)Conduct an empirical study on the complex characteristics of domestic typical urban road network based on MRI system and its adjustment effect on traffic system performanceSupported by the concept of the MRI system,this part first establishes an analysis model of the“adjustment effect of the dual topological network characteristics of urban roads on the performance of the urban road network ”,and then collects 90 typical cities in China(Including refined road network data,related city attribute data and the dynamic Baidu traffic index)as model input.Besides,and the EICN algorithm was also used to construct a dual topology network and MRI system for each urban road network data,and various dual topologies were statistically acquired.Finally,the network structure parameters were combined with the model for exploratory analysis and verification.Compared with the traditional single or a small number of urban system verification analysis,the verification process in this section is the first to achieve a macro analysis of the large sample urban road network dual topology and the overall efficiency of the transportation network.The relevant conclusions quantitatively support the importance of some complex network characteristics for the road network efficiency,such as road betweenness etc.(3)Propose an off-line efficient map matching(EMM)algorithm based on MRI systemIn order to improve the accuracy and speed of off-line map matching for massive GPS trajectory data and meet the growing demand for spatio-temporal data analysis,this section designs an adaptive dynamic segment-wise map matching algorithm(SMRI,Segment-wise matching based on MRI system),and analyzed the level of algorithm efficiency improvement,and finally took Xi'an as an example to verify the actual matching effect of the algorithm.In addition,in order to comprehensively measure the improvement effect of the algorithm,three existing algorithms were also added for comparison.The analysis results show that,compared with the existing mature algorithms,SMRI can improve the matching efficiency by up to 108 times without loss of accuracy.And due to the upgrade of the underlying matching logic,the shorter the GPS sampling interval,the more obvious the matching efficiency improvement.Compared with existing algorithms,this technology can also be combined with MRI systems to greatly popularize various types of urban transportation system research based on high-precision GPS trajectory matching results.The application prospects are broad.(4)Innovatively develop a case study on the spatial and temporal distribution characteristics of urban taxi operating income and detour behavior based on MRI system and EMM technologyThis part firstly combines the EICN algorithm and SMRI algorithm to normalize the matching data of Shenzhen taxis for one week(including the removal of outliers,map matching,MRI scale statistics,etc.),and at the same time combines the taxi rates within the research scope to the taxi revenue.The spatial and temporal distribution was statistically expressed at the scale of the Directional Road Segment(DRS),and the related dual-topology network indexes and other attribute indexes were extracted to establish the Logistic regression model.The DRS variables that affected the income distribution were explored and verified.In addition,this section also designed a reliable detour behavior analysis algorithm using 5 types of travel distances.Through the establishment of a Logistic regression model,the factors that affect the detour behavior was explored and verified by DRS scale.Compared with the characteristics of the traditional vehicle or regional taxi market and operating behavior research based on a more limited view(not effectively combining traffic demand distribution and system structure characteristics),this section of the study is based on the MRI system and ECC technology for the first time at the DRS level.The spatial and temporal distribution of car market and detour behavior were analyzed by spatial statistics and influencing factors.The analysis content covered various traffic network structural parameters and dynamic characteristics.At the same time,combined with the POI location distribution characteristics,it strongly supports relevant research conclusions,taking Shenzhen as an example,an average of 45 completed taxi trips per minute(27.5%)have a detour distance larger than 1.5 kilometers compared with the corresponding shortest path distance.For "manual surveys,small samples,model-oriented " to " automatic data collection,large samples,data-oriented" technology conversion needs,each part of the research content was achieved through multiple case studies and verified a set of MRI establishment methods and EMM technology,and the feasibility of innovative applications in urban traffic system analysis hot issues.This set of research methods based on MRI system and EMM technology can be extended to various large-scale urban traffic system analysis subdivisions.