An intelligent-based prediction system for incident-induced congestion spreading

Traffic congestion is a severe problem in many modern cities around the world. The non-recurring occurrence of incident-induced congestion sometimes may cause large-scale traffic congestion. This paper proposes an intelligent-based prediction system, Fuzzy-Neural Network Prediction System (FNNPS), as decision support to operators. Fuzzy algorithm is used in FNNPS to handle the different sources of heterogeneous input data including the static structure variables of road network, the real-time traffic flow information (i.e. the average volumes and velocity of 5-minute intervals), and incident reports including incident location, type and severity etc. The whole system is linear time complexity to give real-time prediction results. The FNNPS performances show perfect prediction during the congestion spreading process with system parameters of the number of membership functions of inputs and the frequency of re-trainings in a busy area of Beijing.