Traffic flow forecasting based on hybrid deep learning framework

Traffic flow forecasting is a key problem in the field of intelligent traffic management. In this work, we propose a hybrid deep learning framework for short-term traffic flow forecasting. It is built by the multi-layer integration deep learning architecture and jointly learns the spatial-temporal features. According to the highly nonlinear and non-stationary characteristics of traffic flow data, the framework consists of Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs). The former is to capture long temporal dependencies by using Long Short-Term Memory (LSTM) units and the latter is to capture the local trend features. The proposed framework is compared with other traditional shallow and deep learning models for traffic flow forecasting on PeMS datasets. The experimental results indicate that the hybrid framework is capable of dealing with complex nonlinear urban traffic flow forecasting with satisfying accuracy and effectiveness.