An improved method of short-term traffic prediction

Abstract

Traffic congestion is one of the most severely disturbing problems of everyday life in Metropolitan areas. An urban traffic network is at its most vulnerable during peak hours and slight fluctuations of capacity may cause severe congestion. An effective way to avoid this phenomenon is to predict and forestall congestion. Therefore the development of an accurate short-term traffic prediction method could have very real and substantial practical benefits by means of linking it with the urban traffic network control system. In this paper, an improved method is proposed which combines the ANN (Artificial Neural Network) model (used for matching the linear trends) with the fractal self-similar model (used for matching the non-linear trends), and sets proper coefficients of the two results with the real data and real-time. Result shows that the improved method can provide a more accurate prediction for short-term traffic flow in an urban area. For the covering abstract see ITRD E129315.