Traffic flow prediction with strong dynamic spatiotemporal dependence using a hybrid model of transformer and tcn

HIGHLIGHTS

• What: The authors construct Semantic Masking Matrix (𝑀sem): 1 if node 𝑗 is among the top 𝐾 similar nodes to nodes 𝑖 𝑀𝑠𝑒𝑚 (𝑖, 𝑗)={0 otherwise This masking matrix allows the model to focus on node pairs that, despite being geographically distant, exhibit analogous traffic patterns due to similar urban functions. The input reshaping and layer normalization is used to make the model more focused on the spatiotemporal relationships between nodes by treating each time step separately, as follows: 𝑋𝑠𝑝𝑎𝑡=𝑋 resharped to (𝐵 × 𝑁, 𝑇, 𝐶) Then the model will apply layer normalization, as follows: ~ 𝑋𝑠𝑝𝑎𝑡=LayerNorm(𝑋𝑠𝑝𝑎𝑡 ) Then, the Query, Key, and Value Matrices is computed. The model . . .
  

   

  

  If you want to have access to all the content you need to log in!

  Thanks :)

  Username or Email

  Password

  Remember me

  Lost your password?

  If you don't have an account, you can create one here.