Recommendation method for assisting user interest modeling based on using article popularity

The invention discloses a recommendation method for assisting user interest modeling based on using article popularity, which comprises the following steps: S1, dividing an article sequence in a session into a plurality of sample points, inputting the sample points into a graph neural network, forming a graph by each sample point, taking each article as a node, and connecting the sample points according to an article click sequence to form a plurality of edges; continuously updating vector representation of the nodes through information flow of adjacent nodes; s2, inputting vector representations of all nodes in one graph into a sensing network to obtain long-term interests represented by vectors; the vector representation of the last item in the session is regarded as insufficient short-term interest; the article popularity is extracted through the session article interaction information; s3, finally inputting the obtained long-term interests, short-term interests and article popularity into a self-attention