Affiliations: College of Management, Hubei University of Education, Wuhan, China
Correspondence:
[*]
Corresponding author. Lingmin Yang, College of Management, Hubei University of Education, Wuhan, 430205, China. E-mail: andlingmin@21cn.com.
Abstract: In order to overcome the problem of low fitting between traffic uncertainty prediction results and actual values in existing research methods, a traffic flow uncertainty prediction method based on K-nearest neighbor algorithm is proposed. The original database, classification center database, k-nearest neighbor database and intermediate search database are used to construct the database needed in the prediction process. Based on the database, multivariate linear regression is used to assign weights to state variables, and k-nearest neighbor algorithm and Kalman filter are used to update the weights to adapt to the uncertainties of traffic flow until the predicted values are obtained, and the uncertainties of traffic flow are predicted. The experimental results show that the maximum average absolute error and average relative error of the proposed method are 0.018 and 0.02, respectively. Compared with the traditional method, the proposed method has higher overall prediction accuracy, higher fitting degree, and is feasible.
