Traffic Prediction Research For Data Discretizatio

As an infrastructure for social development,transportation facilities have strongly contributed to the rapid development of human civilization.As the scale of transportation networks continues to expand,manual methods of allocating transportation resources can no longer meet the growing demand for high-quality travel.To solve the problem of inefficient and high consumption of manual resource allocation,researchers have introduced intelligent transportation systems.ITS is an efficient management system that integrates advanced sensor technology,computer technology,electronic information technology,and data communication technology with basic transportation facilities and is applied to the transportation field.Traffic prediction is one of the important research directions in intelligent transportation systems.Traffic prediction can provide a data basis for deciding the technical level of highways,road construction planning,and construction of basic transportation facilities.With the rapid development of deep learning methods,the accuracy of traffic prediction models has been improving.However,the outbreak of the new crown epidemic has led to traffic data showing high dispersion and irregularity,and the existing benchmark models cannot accomplish the prediction task of highly discrete traffic data.In this paper,we take the prediction of highly discrete traffic data as the basic research direction and discuss the advantages and shortcomings of the existing related research solutions in depth.For traffic prediction problems in different application scenarios,different data design methods and deep learning models are proposed in this paper,and an innovative data dispersion analysis method is proposed to accomplish the task of predicting highly discrete traffic data.In this paper,the task of predicting highly discrete traffic data is accomplished in a stepwise optimization manner.Firstly,a deep spatio-temporal model is constructed in this paper based on the spatio-temporal correlation of traffic data,to complete the task of capturing the spatio-temporal characteristics of traffic data by the model;secondly,a data dispersion analysis method for highly discrete traffic data is designed in this paper,and the method is used to optimize the deep spatio-temporal model,to improve the model’s ability to characterize highly discrete data;finally,this paper optimises the spatial information storage structure and spatial feature extraction module in the deep spatio-temporal model and combines it with data dispersion analysis methods to accomplish the task of predicting highly discrete traffic data.The innovations and contributions of this paper are as follows.For the problem of spatio-temporal feature mining of traffic data,this paper adopts deep learning modeling to capture the spatio-temporal features of traffic data.On the one hand,this paper proposes a single-layer dual-channel spatial feature extraction module that is used to capture the spatial features and node similarity features of traffic networks;on the other hand,this paper periodically divides traffic data according to their periodic features and designs a temporal feature extraction module to capture the temporal features of traffic data.For the problem of high discrete traffic data prediction,this paper improves the model prediction effect by optimizing the structure of the deep spatio-temporal model.In addition,this paper designs a method for data dispersion analysis based on the deep spatio-temporal model to improve the accuracy of the model for predicting highly discrete traffic data.For the problems of node hierarchy and directional analysis,this paper optimizes the spatial extraction module of the deep spatio-temporal model and introduces data dispersion analysis to accomplish the model’s prediction task for highly discrete traffic data.In this paper,a new spatial information storage structure for a traffic network is designed.Unlike the adjacency matrix that stores two-dimensional planar information,this structure constructs three-dimensional spatial information by overlaying multiple layers of two-dimensional planar information.In addition,this paper designs a new spatial information calculation method to analyze the hierarchy and directionality among nodes in the traffic network.