Research Article

A Dynamic Spatio-Temporal Deep Learning Model for Lane-Level Traffic Prediction

Table 1

Summary of notations.

SymbolDescription

Lane network
Node set and edge set of in
Traffic information of all lanes at timestamp
Traffic information of lane at timestamp
Data-driven adjacent matrix
Distance-based adjacent matrix
Traffic information similarity matrix
Sequence length for input and predict
A constant that controls the contribution of
Degree matrix
Update gate and reset gate in GRU
Cell state and hidden state in GRU
Learnable parameter matrices
Feature fusion gate
Learned spatial features
Learned temporal features
Fused spatio-temporal features