Prediction of New Media Information Dissemination Speed and Scale Effect Based on Large-Scale Graph Neural Network

Algorithm 2

Learning process of the NWIFD model.

	input: cascade graph C, sequence of cascade graph adjacency matrices , time window of observation
	output: predicted information cascade incremental scale
(1)	the Laplacian matrix of the concatenated graph C;
(2)	the graph wavelet for each node ;
(3)	compute the node embeddings of the information cascade graph ;
(4)	Calculate the node embedding of the global graph ;
	Calculate the global matrix
(5)	while not converge do
(6)	Train a bidirectional gated recurrent unit to acquire ;
(7)	for each user in the pair i do
(8)	calculate
(9)	end for
(10)	get ;
(11)	Train a cascaded variational autoencoder to obtain ;
(12)	Obtained by K transformations ;
(13)	Combining sums and sums to make final scale incremental forecasts;
(14)	end while