Research On Key Technologies Of Traffic Trajectories Dataanalysis Based On Artificial Intelligence

Nowadays,with the rapid development of location technology,traffic trajectory data presents exponential growth,and the data contains the spatio-temporal characteristics and behavioral information of moving objects.It has both theoretical value and practical demand to analyze the information hidden behind the massive traffic trajectory.With the acceleration of urbanization in China,the problems brought by urbanization have become increasingly prominent,such as traffic congestion and unreasonable distribution of urban public infrastructure.Traffic trajectory data records residents’ travel trajectory.Through the analysis of artificial intelligence technology,it can reflect residents’ activity rules and spatial distribution of urban hot spots,providing a reference for solving the problems caused by urbanization.Is wisdom city track traffic data analysis,intelligent transportation in the future urban road planning and layout of the main research fields,but at this stage,track traffic data analysis application also exist some problems in the actual scene,first of all,in the trajectory data preprocessing,positioning device access to data and there is a deviation between the real path problem,Due to the increasing complexity of the road,the original data processing technology has been unable to meet the practical needs.Secondly,the existing trajectory data flow prediction model has low prediction accuracy and urgently needs to be improved.Thirdly,the existing track data hot spot mining technology has been unable to meet the needs of scientific research in the face of the computation efficiency of large-scale track data.Therefore,this paper uses the trajectory data of ride-hailing from Didi to carry out research from the perspective of trajectory data preprocessing technology,trajectory data pedestrian flow prediction technology and traffic trajectory data hot spot analysis technology.The main research contents and achievements in this paper are as follows:(1)In view of the deviation between the trajectory data and the real path,this paper proposes a map matching algorithm based on the HMM model.This algorithm starts from the global route,selects K candidate paths which are closest to the data to be processed,finds out the candidate points through the candidate paths,and generates the state transition matrix by using the normal distribution.The state transition matrix is generated by the change of speed,Angle and other information,and the trajectory points are finally matched to the real line.The algorithm is compared with other algorithms,and the matching accuracy is improved obviously.(2)for trajectory data of residents travel flow prediction model based on the problem of low accuracy,difficulty for traffic forecast model is to consider the time complexity and space complexity and time complexity refers to: time dimension flow factors may affect the residents travel,such as weather,temperature,space complexity refers to: the influence of adjacent or no space between adjacent.Paper first trajectory data analysis through the residents travel intensity,time and space distribution and time loss situation,by means of correlation analysis of the factors that influence the residents travel secondly divided to predict the area,to analyze the mutual influence between,is put forward based on the LSTM figure convolution of network,the network through the extraction of a neural network for time information,Finally,the information extracted from the two networks is put into an LSTM for training and prediction.The experimental comparison shows that the algorithm in this paper is better than the existing network model.(3)for trajectory the hot issue within the area of data mining based on the problem of low efficiency of calculation,this paper proposes a hot based on DBSCAN clustering algorithm,the algorithm based on the confidence level in a lot of trajectory data and the KL divergence as the standard to select the more important data,and caused a lot of useless information for computation of the waste of time,The function of automatic parameter selection is realized,and the clustering result is compared with the heat map and the real POI information,which highlights the superiority and correctness of the algorithm.