Understanding Dynamic Spatio-Temporal Contexts in Long Short-Term Memory for Road Traffic Speed Prediction
This work addresses traffic flow prediction for intelligent transportation systems, but it is incremental as it builds on existing LSTM methods with localized dynamic weights.
The study tackled the problem of predicting road traffic speed by addressing the lack of dynamic spatio-temporal interactions in existing models, proposing a dynamically localized LSTM that incorporates spatial and temporal dependencies, and achieved superior prediction performance compared to baseline methods.
Reliable traffic flow prediction is crucial to creating intelligent transportation systems. Many big-data-based prediction approaches have been developed but they do not reflect complicated dynamic interactions between roads considering time and location. In this study, we propose a dynamically localised long short-term memory (LSTM) model that involves both spatial and temporal dependence between roads. To do so, we use a localised dynamic spatial weight matrix along with its dynamic variation. Moreover, the LSTM model can deal with sequential data with long dependency as well as complex non-linear features. Empirical results indicated superior prediction performances of the proposed model compared to two different baseline methods.