[Paper Review] Deep Bidirectional and Unidirectional LSTM Recurrent Neural Network for Network-wide Traffic Speed Prediction

Zhiyong Cui, Ruimin Ke+2 more|arXiv (Cornell University)|Jan 7, 2018

Traffic Prediction and Management Techniques44 references397 citations

TL;DR

The paper proposes a deep stacked bidirectional and unidirectional LSTM (SBU-LSTM) architecture to predict network-wide traffic speed, leveraging bidirectional temporal dependencies and a masking mechanism to handle missing data.

ABSTRACT

Short-term traffic forecasting based on deep learning methods, especially long short-term memory (LSTM) neural networks, has received much attention in recent years. However, the potential of deep learning methods in traffic forecasting has not yet fully been exploited in terms of the depth of the model architecture, the spatial scale of the prediction area, and the predictive power of spatial-temporal data. In this paper, a deep stacked bidirectional and unidirectional LSTM (SBU- LSTM) neural network architecture is proposed, which considers both forward and backward dependencies in time series data, to predict network-wide traffic speed. A bidirectional LSTM (BDLSM) layer is exploited to capture spatial features and bidirectional temporal dependencies from historical data. To the best of our knowledge, this is the first time that BDLSTMs have been applied as building blocks for a deep architecture model to measure the backward dependency of traffic data for prediction. The proposed model can handle missing values in input data by using a masking mechanism. Further, this scalable model can predict traffic speed for both freeway and complex urban traffic networks. Comparisons with other classical and state-of-the-art models indicate that the proposed SBU-LSTM neural network achieves superior prediction performance for the whole traffic network in both accuracy and robustness.

Motivation & Objective

• Motivate deep learning for traffic forecasting across large networks and complex urban environments.
• Develop a neural architecture that captures both forward and backward temporal dependencies in traffic data.
• Enable robust predictions even with missing input values via masking mechanisms.
• Show scalability of the model to network-wide prediction and different traffic network types.

Proposed method

• Introduce a deep stacked bidirectional and unidirectional LSTM (SBU-LSTM) architecture for traffic speed prediction.
• Use BDLSM layers to capture spatial features and bidirectional temporal dependencies from historical data.
• Incorporate a masking mechanism to handle missing input values during training and inference.
• Apply the model to both freeway and complex urban traffic networks.
• Compare against classical and state-of-the-art models to assess accuracy and robustness.

Experimental results

Research questions

• RQ1 Can a deep BD-LSTM-based architecture improve network-wide traffic speed prediction over traditional models?
• RQ2 Does combining bidirectional temporal dependencies with unidirectional layers enhance forecasting accuracy and robustness?
• RQ3 How effectively does masking handle missing data in traffic speed inputs?
• RQ4 Is the proposed model scalable to large networks spanning freeway and urban contexts?

Key findings

• The SBU-LSTM architecture achieves superior prediction performance for the entire traffic network compared to reference models in terms of accuracy and robustness.
• Bidirectional temporal dependencies help capture backward influence in traffic speed data.
• The masking mechanism enables the model to handle missing input values effectively during training and prediction.
• The approach is applicable to both freeway and complex urban traffic networks.
• The deep architecture demonstrates improved performance and resilience over traditional approaches.