Research And Application Of Urban Traffic Big Data Basedon Machine Learning

As transportation becomes more and more developed and brings convenience to people,it also brings problems such as traffic congestion and urban safety.For the acceleration of my country’s traffic informatization process,it provides the possibility of analyzing people’s travel characteristics from massive traffic data and constructing prediction models to solve traffic problems.This thesis uses machine learning technology to study the laws that exist in traffic data,and provides reference data for people’s travel,city planning,and tourism bureau management.The main sources of data are bus IC card data and mobile phone signaling data,which have strong application value.Use machine learning algorithms to conduct targeted design and research and analysis of traffic data.(1)The thesis first is the research and analysis of the power rate distribution based on IC card data.First,it analyzes the characteristics and distribution of IC data,proposes a non-limiting substation algorithm,and obtains the time interval between stations,and then the data from two different regions are used to verify the availability of the non-limiting substation algorithm.Then analyze the obtained time interval between the bus station and the station.It is found that it has a heavy tail phenomenon,which is in line with human behavior dynamics.And through the machine learning linear regression algorithm to model the time interval data,it is concluded that the station-to-station time interval is subject to power law distribution.The analysis of IC card data in this paper lays a theoretical foundation for bus scheduling.(2)Then the paper studies and analyzes the data of mobile phone signaling,and proposes an adaptive density clustering algorithm based on confidence to study the spatial and temporal distribution characteristics of mobile phone signaling.And takes tourists from other provinces in coastal areas as the main research objects,through the adaptive density clustering algorithm based on confidence and the traditional density clustering algorithm for comparison and analysis,and the thermal map obtained by the visualization tool for reference analysis to illustrate the accuracy and superiority of the algorithm.The purpose of this chapter is to use mobile phone signaling data to dig out the main activity locations of tourists in different regions during the peak tourism period of a Golden Week in a coastal city,and obtain the main distribution locations of traffic flow.(3)At last,the thesis studies the forecast of people flow based on mobile phone signaling data,and analyzes the data of tourist cities in the golden week of tourism(Eleven Golden Week)according to the distribution of people flow in different regions.Then use the LSTM algorithm and ARIMA algorithm to predict the data sets of different regions,and prove the efficiency and availability of the LSTM prediction algorithm to predict the tourist flow in different regions of the tourist city during the Golden Week.It predicts the trend of people flow in different regions of the tourist city in a certain period of time in the future,provides a reference program for tourists,and also provides beneficial help for traffic management and public safety in different regions of the tourist city,making the mode of transportation from real-time monitoring to predictable future.The paper proposes non-limiting substation algorithm,linear regression algorithm,adaptive density clustering algorithm based on confidence,LSTM algorithm,and then demonstrates through experiments to prove the feasibility and effectiveness of the proposed algorithm for traffic data mining.In urban planning,public safety and traffic management have certain research value and application value.