Search

Abstract Forecasting the short-term speed of moving vehicles on roads plays a vital role on traffic control and trip planning, which however still remains a challenging task when the high accuracy is required. In this paper, we propose a novel approach to the short-term traffic speed forecasting, which takes into account the influence of different traffic attributes, such as traffic flow, traffic speed, road occupancy and traffic density, of adjacent roads on the traffic speed. In addition, in order to obtain the more accurate relation between traffic speed and traffic attributes, we employ the idea of piecewise correlation function and adopt the Jenks clustering method with dynamic programming to determine the segment intervals of relation. We validate our approach based on the real data collected from Wenzhou and Hangzhou, two large cities located in eastern China. The extensive experimental results show that, compared with the state-of-the-art approaches, our approach has the higher stability and accuracy, especially for 5-minute and 10-minute speed prediction. Highlights • We consider only traffic attributes most relevant to speed and also their weights. • We consider the differences of impacts on speed among adjacent roads. • We employ piecewise correlation function between traffic speed and attributes. • We adopt Jenks clustering with dynamic programming to determine segment intervals. • We validate our approach based on the real data collected from two large cities. [ABSTRACT FROM AUTHOR]