SBAG: A hybrid deep learning model for large scale traffic speed prediction

Abstract

Intelligent Transportation System (ITS) is the fundamental requirement to an intelligent transport system. The proposed hybrid model Stacked Bidirectional LSTM and Attention-based GRU (SBAG) is used for predicting the large scale traffic speed. To capture bidirectional temporal dependencies and spatial features, BDLSTM and attention-based GRU are exploited. It is the first time in traffic speed prediction that bidirectional LSTM and attention-based GRU are exploited as a building block of network architecture to measure the backward dependencies of a network. We have also examined the behaviour of the attention layer in our proposed model. We compared the proposed model with state-of-the-art models e.g. Fully Convolutional Network, Gated Recurrent Unit, Long-short term Memory, Bidirectional Long-short term Memory and achieved superior performance in large scale traffic speed prediction.