Research On Urban Road Traffic Status Prediction Based On GPS Data Of Online Ride-hailing

The frequent occurrence of urban road traffic jams and accidents has brought great trouble to people’s traffic and social production.Although many city managers have made many different attempts to solve the current problems,these phenomena still exist in various cities and show an upward trend.How to effectively improve the operation efficiency and management of urban roads under the existing traffic facilities Management ability is the main problem in the field of transportation.In the urban road network,car Hailing is running all the time,including a large number of comprehensive urban traffic space-time information,and the GPS data has the characteristics of high reliability and large amount of data.The collected data is transmitted to the control center of each major platform through the wireless network.The unified management of the platform ensures the integrity and real-time of the data,so the GPS data of car Hailing is very suitable for urban roads Through the prediction of traffic state,the traffic management department can make more targeted traffic control and guidance measures,so that travelers can plan their travel plans in advance and avoid congested areas or periods in time,so as to reduce the traffic pressure of urban roads.In this paper,the data of car hailing in the study area are processed and analyzed,and the historical traffic information is mined.Combined with the relevant algorithms,the traffic state discrimination and traffic flow parameter prediction are studied.This paper first introduces the content of online car Hailing data in detail,describes the information of the study area,divides the road sections and selects the statistical cycle,and describes the problems of the original data,such as the confusion of recording time,the lack of data,the drift of positioning data,etc.,and carries out the data preprocessing,and then analyzes the traffic flow parameters,according to the traffic flow parameters selection In principle,the average speed,traffic flow and travel time index are selected as the traffic characteristic parameters,and the data are analyzed from three aspects: the spatial distribution of traffic congestion,the time distribution of online car Hailing travel demand and the degree of traffic congestion in history.After the above processing,the second-order clustering method is used to obtain the Schwarz Bayes criterion and other indicators under different cluster number values to determine the optimal cluster number.The traffic state is divided into four types: smooth,basically smooth,slow and congested.After extracting the relevant parameters,the traffic state discrimination model based on GA-FCM algorithm is constructed.The model is found through experiments Excellent performance in traffic identification.In addition,this paper proposes a traffic flow parameter prediction model based on cnn-lstm.Pearson correlation test is used to analyze the correlation of traffic flow parameters from the two dimensions of time and space,and the corresponding influencing factors are extracted to construct the model input feature vector.Then other models such as ARIMA and Gru are simulated.Through comparison,it is found that the prediction effect of this model is better.Finally,combined with the traffic state discrimination model and traffic parameter prediction model,the overall technical scheme of the traffic state prediction system is discussed in detail.