Short-term traffic flow prediction with nearest trajectory segments

Abstract

As a key technology of Intelligent Transportation System(ITS), short-term traffic flow prediction is fundamental to traffic control and management. This paper proposes a prediction method based on nearest trajectory segments in reconstructed phase space. First, phase space reconstruction is introduced to recover dynamics traffic flow time series. Then a optimized metric which integrates Euclidean distant and cosine similarly of trajectory segments is proposed to select nearest trajectory segments in phase space. Finally, the predicted traffic flow value is obtained from the predicted vector computed with nearest trajectory segments. Case study with traffic flow data collected from Guangshen Freeway proves prediction accuracy.