Research On Short-term Forecast Method Of Freeway Vehicle Speed Under Rainy Weather

The short-term prediction of freeway speed is an important basis for highway traffic management.The speed of freeway vehicles is easily affected by accidents,weather,and other random factors,that has strong time-variability and randomness.These increase the difficulty of traffic forecasting.This paper focuses on the impact between rainfall weather and highway speeds,fully explores the influence between the weather data and speed data which from the traffic survey system,analyzes the influence of rainfall weather on the speed of freeway traffic flow,and then establishes a short-term prediction model of freeway speed under the condition of rainfall in order to improve the accuracy and robustness.First of all,this paper preprocesses the average driving speed data and the rainfall data during the same time period with the traffic data,extracts the flow and speed information of each vehicle type,and analyzes the variation regularity of each type vehicles’ speed;then,this paper systematically analyzes the factors affecting the driving speed and then eliminates the influence of other factors.It focuses on analyzing the correlation between the rainfall factors and the vehicle speed,and further analyzes the effect of rainfall weather on the speed of the highway.Based on the analysis of the influence of rainfall conditions on freeway speed,a short-term prediction model of speed,Bayes-KNN algorithm,is used to identify the similar speed mode and predict the speed which based on pattern recognition algorithm is proposed.This algorithm effectively combines the advantages of the Bayesian classification algorithm and the K-nearest neighbor prediction algorithm,and improves the K-nearest neighbor algorithm’s prediction efficiency in the big data environment.The feature vector of the algorithm includes rainfall factors,which can directly search the historical average driving speed under the sub-database which weather conditions are most similar to the current time period,and the robustness of the algorithm is effectively improved because of not affected by random events.Finally,the feasibility of the Bayes-KNN prediction algorithm was verified based on the speed data from the section of the Ninghu Expressway from Suzhou to Wuxi and the rainfall data from the surface monitoring station.And the average absolute error,average relative error,mean square error,and mean square relative error were evaluated.