The Forecasting of Urban Transportation: A Neural Network Approach

Abstract In this paper, the neural network models are employed in making forecasts of urban transportation. A special neural network architecture with feedback nodes (FBNN) is developed for this purpose. Comparing with the typical time series forecasting method, the neural network approach is more effective and precise in multi-variable forecasting. The neural network approach also provides a practical way to select and analysis relative factors in a reasonable manner which is especially important in building nonlinear models and difficult in general statistical methods. In addition, the neural network approach is capable of combining different categories of data into an integrative model, which improves the effectiveness of forecasting.