Attention-Based Multi-fusion Method for Citation Prediction
As the most common research style in the process of academic exchange, the paper plays a vital role in the specific links of knowledge communication, academic cooperation, and scientific research education. However, in the traditional field of bibliometrics, how to quantitatively evaluate the influence of a paper generally depends on the number of citation as a reference standard. The number of citation is an important indicator for evaluating papers, and it has serious problems of lag. Therefore, based on the relevant meta-information generated in the publication process of the literature, the prediction of the future influence of the literature can make up for the above defects. In order to accurately predict the future citations of the paper, this paper constructs the Attention Convolution Neural Network model and combines the bibliometrics and alternative metrology-related features to enrich the input vector. Experiments on data sets collected from WOS and ResearchGate show that the model has improved accuracy compared to traditional prediction models.
KeywordsAttention convolution neural network Bibliometrics Altmetrics
- 5.Gazni, A., Didegah, F.: Investigating Different Types of Research Collaboration and Citation Impact: A Case Study of Harvard University’s Publications. Springer-Verlag, New York (2011)Google Scholar
- 13.Vanclay, J.K.: Factors affecting citation rates in environmental science [J]. J. Inf. 7(2), 265–271 (2013)Google Scholar
- 17.Yan, R.,Tang, J., Liu, X., et al.: Citation count prediction: learning to estimate future citations for literature. In: Proceedings of the 20th ACM International Conference on Information and Knowledge Management, pp. 1247–1252. ACM (2011)Google Scholar
- 18.Mistele, T., Price, T., Hossenfelder, S.: Predicting citation counts with a neural network [J]. arXiv preprint arXiv:1806.04641 (2018)
- 19.Abrishami, A., Aliakbary, S.: NNCP: a citation count prediction methodology based on deep neural network learning techniques [J]. arXiv preprint arXiv:1809.04365 (2018)