Combined Neural Network Approach for Short-Term Urban Freeway Traffic Flow Prediction

Abstract

Short-term traffic flow prediction is an essential component of urban traffic management and information systems. This paper presents a new short-term freeway traffic flow prediction model based on combined neural network approach. This model consists of two modules: a self-organizing feature map neural network and an Elman neural network. The former classifies the traffic conditions in a day into six patterns, and the later specifies the relationship between input and output to provide the prediction value. The inputs of the Elman neural network model include three kinds of data: the several time series of the prediction location, the historic data of the predictive time interval in the same weekdays, the time series of the adjacent location. The performances of the combined neural network model are validated using the real observation data from 3^rd ring freeway in Beijing.