Research On Road Congestion Prediction Based On Traffic Flow Similarity And Difference

In recent years,the world economy has developed rapidly.However,in many countries,the increase of transportation facilities is slower than the increase of motor vehicles,resulting in more and more serious traffic congestion.Traffic congestion will cause significant time waste and economic waste,as well as air pollution,noise pollution,etc.Therefore,it is necessary to study how to ease traffic congestion.With the advancement of technology and the improvement of hardware,data science has become more and more widely used,so researchers have tried to use data science to alleviate traffic congestion,of which traffic flow prediction and traffic congestion prediction are the most important ways.However,most researchers who use traffic flow sequences pay attention to the original traffic flow sequence directly,and do not take into account the similarity and difference of traffic flow.We find that the use of these two properties can improve the accuracy of the model,so the traffic congestion prediction based on traffic flow similarity and difference is carried out in this paper.In this paper,the similarity and difference of traffic flow data are analyzed by data first.Then,using these two properties,this paper proposes a "Congestion Assessment Model based on Traffic Flow Change Score" and a "Traffic Congestion Prediction Model Based on Offset and Split Prediction Algorithm”.In the "Road Condition Assessment Model based on Traffic Flow Change Score",the traffic flow change score is defined as an indicator to measure the degree of change of traffic flow.Based on the similarity and difference of traffic flow,this model establishes traffic flow forecast model for traffic flow first.Then predictive models and a probability model are combined to analyze the change score of traffic flow.Finally,the status of the road is evaluated according to the change score."Traffic Congestion Prediction Model Based on Offset and Split Prediction Algorithm" is divided into two parts.The first part is "offset and split prediction algorithm".The offset and split prediction algorithm is a traffic flow prediction algorithm.It splits traffic flow into a basis sequence representing similarity and a differential sequence representing difference first.Then offset values and these two sequences are used to predict traffic flow.The second part is a method to judge the road network congestion state suitable for this model.This model uses clustering method to label the historical data first.Then it uses the labeled data to train a classifier.Finally this model uses the classifier to judge the traffic congestion state of the predicted traffic flow.Combining these two parts,this model can predict the target traffic flow and determine its road network congestion status.Finally,this paper verifies the validity of the two models on a real data set.The experimental data set used is the traffic data of each road network node in the month of May 2013 in a certain urban area in China.The daily data volume exceeds 5 million.The Road Condition Assessment Model based on Traffic Flow Change Score can determine the reasonable congestion time and congestion time in the real data,so as to evaluate the congestion state of road network,and prove the validity of the model.For the Traffic Congestion Prediction Model Based on Offset and Split Prediction Algorithm,this paper first proves the accuracy of the offset split prediction part through comparative experiments.Next,this model predicts a relatively accurate traffic flow sequence on the real traffic flow data and judges the congestion zone that meets common sense.This shows that the traffic congestion prediction model can effectively predict traffic flow and traffic congestion.