How to Identify Patterns of Citywide Dynamic Traffic at a Low Cost? An In-Depth Neural Network Approach with Digital Maps

Figure 7

Vectors captured by the proposed stacked convolutional autoencoder. The latent space features are three dimensions. Vectors are divided into four categories, morning and evening peak hours and in and out of the restriction period. Scatter plots are organized by peak hours: (a) morning peak hours and (b) evening peak hours. The red circle marker is vector in the restriction period, and the blue star marker is vector out of the restriction period.

(a)

(b)