The citation network is a complex network that describes the scientific research results of scholars, the development of scientific fields, and the relationships between various academic fields. In the exploration and analysis of citation networks in actual scenarios, the data in the citation network is often incomplete or contains noise. Therefore, the purpose is to determine and predict whether there is an interaction or relationship between two documents in the citation network. The citation link prediction task has very significant research value. Among them, the link prediction method based on graph neural network has developed rapidly due to its excellent graph structure learning ability. However, there are problems in traditional citation link prediction methods based on graph neural networks that cannot efficiently utilize document attribute features and are similar to the cold start problem in recommendation systems. Aiming at these problems that arise in current graph neural network methods, this paper proposes a graph neural network link prediction model Warmer-GNN based on literature semantic information enhancement (Warmer refers to the problem of solving the cold start problem, and GNN refers to the graph neural network) . The model first establishes the document attribute feature relationship graph and the document citation relationship graph from the perspective of semantic information enhancement, then obtains the information of nodes in the feature map and citation map through the self-attention mechanism, and finally uses a mixture of positive and negative samples to target the feature map Perform negative sampling optimization.

Yi Chen. Cold Start Problem in Citation Link Prediction. Academic Journal of Computing & Information Science (2024), Vol. 7, Issue 5: 116-122. https://doi.org/10.25236/AJCIS.2024.070515.

[1] Gilmer J, Schoenholz S S, Riley P F, et al. Neural message passing for quantum chemistry[C]. International conference on machine learning. PMLR, 2017: 1263-1272.

[2] Botev Z I, Grotowski J F, Kroese D P. Kernel density estimation via diffusion[J].Ann Stat 2010, 38(5): 2916-2957

[3] Perozzi B, Al-Rfou R, Skiena S. Deepwalk: Online learning of social representations[C].Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. 2014: 701-710.

[4] Hamilton W, Ying Z, Leskovec J. Inductive representation learning on large graphs[J]. Advances in neural information processing systems, 2017, 30.

[5] Qian T, Liang Y, Li Q. Solving cold start problem in recommendation with attribute graph neural networks [J]. arXiv preprint arXiv:1912.12398, 2019.

[6] Huang T, Dong Y, Ding M, et al. Mixgcf: An improved training method for graph neural network-based recommender systems[C].Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. 2021: 665-674.

[7] Sen P, Namata G, Bilgic M, et al. Collective classification in network data[J]. AI magazine, 2008, 29(3): 93-93.

[8] Kipf T N, Welling M. Semi-supervised classification with graph convolutional networks[J]. arXiv preprint arXiv:1609.02907, 2016.

[9] Kipf T N, Welling M. Variational graph auto-encoders[J]. arXiv preprint arXiv:1611.07308, 2016.

[10] Veličković P, Cucurull G, Casanova A, et al. Graph attention networks [J]. arXiv preprint arXiv:1710.10903, 2017.