Fig. 4
From: Multi-scale spatio-temporal graph neural network for urban traffic flow prediction
Sensitivity of key model parameters: (a) influence of the number of layers in the STBlock module on model accuracy; (b) influence of the number of Chebyshev polynomial coefficients on model accuracy; (c) influence of prediction horizon on model accuracy; (d) influence of the size of the convolution kernel in the graph convolutional network on model accuracy.
