Research On Efficient Prediction Method For Urban Checkpoint Traffic Flow

The urban traffic flow forecast can help the traffic control department to conduct congestion relief and provide route planning suggestions for travellers.How to make accurate and efficient traffic flow prediction is a very practical task.Regarding traffic flow forecasting,there have been many fruitful results at home and abroad,but the existing methods still have certain limitations:(1)The different regions of the forecasting model result in poor scalability.For example: a model that predicts data such as vehicle speed,traffic flow,and road occupancy rate obtained from traffic sensors has high prediction accuracy on suburbs and highways,but it is not suitable for urban areas with complicated road conditions and heavy traffic;(2)Model training takes a long time.In recent years,the prediction method based on neural network design requires a lot of data to train and adjust the model while achieving high accuracy,and the parameters of the model may change over time.The model needs to be retrained to improve the accuracy of the prediction.The resulting long time consumption makes the prediction system lack real-time performance and cannot achieve real-time prediction;(3)The prediction model's real-time prediction accuracy under complex road networks is not high.Some prediction methods only consider the traffic time series data,and ignore the spatial traffic flow of the neighboring checkpoint,resulting in the model's prediction accuracy is not high when the traffic flow changes sharply.In view of the above problems,this paper studies the real-time prediction method of urban road checkpoint from two different dimensions of time series and space,and on this basis,develops a set of prototype system for traffic flow prediction of urban checkpoint.This paper summarizes,sorts and analyzes the relevant methods of traffic flow forecasting at home and abroad,and completes the following work:(1)Processing and feature extraction of checkpoint traffic data,and using the random forest model in machine learning to extract the features to be tested.Based on the experimental analysis of real urban road traffic flow data,when using the extracted feature values??of the checkpoint data(time point features,holiday features,time period periodic features,time stage features and surrounding checkpoint features,etc.),through the random forest model the accuracy of forecasting the traffic volume in the next 5minutes can reach 83%,which is 6% higher than the commonly used ARIMA model.It solves the problem that the above model is not applicable to urban areas and the prediction accuracy of complex networks is not high.(2)As a parameter learningmethod,random forests need to be retrained over time to maintain the prediction accuracy of the model parameters,resulting in long time-consuming and the model cannot make real-time predictions.As a non-parametric prediction model,KNN can avoid this problem.Therefore,this paper designs and implements a traffic flow prediction model based on the non-parametric regression KNN model.This model can obtain accurate and efficient traffic flow prediction by considering K historical values??by considering the correlation of checkpoint data in time and space.Based on the experimental analysis of real urban road traffic data,the short-term prediction of the model in 5 minutes takes 1.3 seconds,and the prediction accuracy reaches 91%.It breaks through the three limitations of the traffic prediction model,which is not universal,the training takes a long time,and the accuracy is poor.(3)This paper designs and implements a prototype system for traffic flow prediction for urban road crossings,including a data processing module,a data access module,and a flow prediction module.The system is implemented on the basis of the KNN prediction model designed in this paper.First,it processes and stores urban road checkpoint data in real time.Secondly,obtain relevant data from the historical database according to the customer's predicted demand,and finally make an accurate and efficient urban checkpoint traffic flow forecast.