| Popis: |
For more than 40 years, various statistical time series forecasting, and machine learning methods have been applied to predict the short-term traffic flow. More recently, deep learning methods have emerged to show better results for short-term traffic flow prediction. For multi-step-ahead prediction, researchers have used iterative (also known as recursive) and direct (also known as independent) strategies with statistical methods for preparing input data, building models and creating forecasts. However, the iterative and direct strategies are not combined with the recurrent neural network architectures. Hence, we present the impact of these two strategies on accuracy of the Recurrent Neural Network models for short-term traffic flow forecasting. |
