Travel Time Prediction Of Expressway Using K Nearest Neighbor Method

With the continuous increase in the number of motor vehicles,the problem of congestion on expressway has become increasingly severe.The environmental deterioration,the decline in travel quality,and traffic accidents caused by traffic congestion have also increased.The information service level of expressway has been put forward with higher requirements.and travel time forecasting is the basis for realizing traffic control and guidance,which can alleviate traffic congestion to a certain extent.Therefore,this article studies the travel time prediction of expressway to obtain more accurate prediction results.Several classical prediction methods are summarized,in order to select prediction methods for fitting research objects.Considering the accuracy and speed requirements of the prediction of expressway travel time,and the data-driven characteristics of the K Nearest Neighbor Method more accord with existing data,in this paper we choose K Nearest Neighbor Method to predict,then analyze the five key steps of K Nearest Neighbor Method in detail.In order to implement the proposed improvement method,the following works have been done:1)Improvement of the state vector.We use principal component analysis to reduce the number of state vectors and contain enough feature information as well.2)Improvement of historical database.Combine the toll data and traffic incident data to establish sub-databases under traffic incidents,and use Bayesian classification to establish sub-databases under normal traffic conditions.The structure of the database is optimized,and the complexity of searching the nearest neighbor in the historical database is reduced and the efficiency of the operation is improved.3)Improvement of K nearest neighbor search.According to the classification result of the historical database,use different K values in different sub-databases,so variable K search is achieved.To prove the effectiveness of the improvement and obtain more accurate prediction results,taking the Guangzhou Airport Expressway South Line as an example,selects the representative within 10 days of the toll data and traffic incident data,the toll data includes information about the toll stations and time of vehicles entering and leaving the toll station,the traffic event data includes information about the location and time of accidents and construction.The effectiveness of the proposed K Nearest Neighbor Method is verified,and it can be proved that the accuracy of the prediction and the computational efficiency can be improved whether it is applied separately or jointly.