Short-Time Traffic Forecasting of Urban Road Network: An ANN Model Based on DTW Clustering

Fast and accurate short-term traffic volume prediction is the core content of ITS. However, the traditional prediction methods for a certain road section cannot provide useful information for urban traffic management. This study proposes a method to predict the traffic flow distribution of the road network for a period of time in the future. K-means clustering with dynamic time warping algorithm (DTW) is used to identify the characteristics of time series of traffic flow. Artificial neutral network (ANN) is trained to predict traffic flow of the whole network. The short-term traffic flow prediction model of urban road network based on DTW and ANN is established and compared with the ARIMA, the ANN model, and the emerging LSTM. DTW-ANN model has better prediction effect in data sets of various intervals and can effectively predict the overall traffic distribution of urban road network.