A 2-Layered Graph Based Diffusion Approach for Altmetric Analysis

The research shared on a digital social media has enabled us to measure the impact of academic entities beyond the conventional bibliometric community. We explored a diffusion-based metrics to measure the influence of academic entities in social media using 2-layered graph where the first layer is the graph between academic and social media entities and a second layer is the graph between social media entities. We employed the heat diffusion algorithms to measure the social impact of academic entities and evaluate them by (i) predicting links between academic entities and social media and (ii) suggesting memes for the academic entities. Our analysis on predicting links between scientist and social media entities showed the AUC-ROC score of 0.73 and the AUC-PR score of 0.30. Similarly, predicting links between scientific publications and social media entities showed the AUC-ROC score of 0.80 and the AUC-PR score of 0.19. Our approach also provides decent social media entities (memes) suggestion for scientific publications.