K-Nearest Neighbor Model Based Short-Term Traffic Flow Prediction Method

Abstract

Forecasting short-term traffic flow is a key issue in the field of Intelligent Traffic System (ITS). Most of the existing researches are oriented to urban road network, the development of urban traffic flow forecasting technology provides a good reference for traffic flow forecasting in the port. However, the complexity and uncertainty of port traffic determine that it is necessary to study the forecasting methods of port traffic flow pertinently. Based on the K- Nearest Neighbor (KNN) algorithm, this paper forecasted short-term traffic flow, and defined the factors affecting short-term traffic flow in the port, such as port working status and operation instruction characteristics, as state vector, and used K-Dimension Tree (KD tree) to reduce the time complexity of neighbor searching. One-month traffic flow data of an inland port in Chongqing was selected to verify the prediction algorithm. Firstly, the number of neighbors was identified, and 8 was the best, then the prediction accuracy was verified to meet the usage requirements, which could provide reference for scientific planning and management of port traffic.