Despite the rise of multilayer networks and their applications for the real world systems, the problem of link prediction is still one of the toughest to address. In this paper, we investigate the problem of link prediction in the multilayer scientific collaboration network. Our proposed solution alters the classic stacking technique for the supervised link prediction in terms of distribution of the training and testing data according to the structure of a multilayer network with training number of models for each layer to predict link formation in a target network. Experimental results show that our approach has positive effect on the link predictions quality, nevertheless, the influence of non-target layers on the resulting prediction is moderately low.
CITATION STYLE
Mbogo, G. K., Visheratin, A., & Rakitin, S. (2018). Layer-wise model stacking for link prediction in multilayer networks. Case of scientific collaboration networks. In Studies in Computational Intelligence (Vol. 689, pp. 117–126). Springer Verlag. https://doi.org/10.1007/978-3-319-72150-7_10
Mendeley helps you to discover research relevant for your work.