Community-guided link prediction in multiplex networks

Machine learning prediction of academic collaboration networks

We investigate the different roles played by nodes' network and non-network attributes in explaining the formation of European university collaborations from 2011 to 2016, in three European Research Council (ERC) domains: Social Sciences and Humanities (SSH), Physical and Engineering Sciences (PE), Life Sciences (LS), as well as multidisciplinary collaborations. On link formation in collaboration networks, existing research has not yet compared and simultaneously examined both network and non-network attributes. Using four machine learning predictive algorithms (LASSO, Neural Network, Gradient Boosting, and Random Forest) our results show that, over various model specifications: (i) best model link formation accuracy is larger than 80%, (ii) among the non-network attributes, public funding plays an important role in PE and LS, (iii) network attributes count more than non-network attributes for the formation, sensibly increasing accuracy, (iv) feature-importance scores show a different ordering in the four domains, thus signalling different modes of knowledge production and transmission taking place within these different scientific communities.

Impact of Centrality Measures on the Common Neighbors in Link Prediction for Multiplex Networks

Complex networks are representations of real-world systems that can be better modeled as multiplex networks, where the same nodes develop multi-type connections. One of the important concerns about these networks is link prediction, which has many applications in social networks and recommender systems. In this article, similarity-based methods such as common neighbors (CNs) are the mainstream. However, in the CN method, the contribution of each CN in the likelihood of new connections is equally taken into account. In this work, we propose a new link prediction method namely Weighted Common Neighbors (WCN), which is based on CNs and various types of Centrality measures (including degree, k-core, closeness, betweenness, Eigenvector, and PageRank) to predict the formation of new links in multiplex networks. So, in this model, each CN has a different impact on the node connection likelihood. Moreover, we investigate the impact of interlayer information on improving the performance of link prediction in the target layer. Using Area under the ROC Curve and precision as evaluation metrics, we perform a comprehensive experimental evaluation of our proposed method on seven real multiplex networks. The results validate the improved performance of our proposed method compared with existing methods, and we show that the performance of proposed methods is significantly improved while using interlayer information in multiplex networks.

Analyzing the Evolution of Interdisciplinary Areas

Recently, various new areas of research have been of great interest to researchers. As these areas are highly based on academic and industrial needs, it is necessary to examine the change and evolution in research. This study proposed a framework for identifying emerging areas and their evolution. The proposed framework suggests that latent Dirichlet allocation is applied to identify emerging topics and their networks in such interdisciplinary areas. The simulation for empirical network analysis was then applied to the identified topic networks to terminate continuous evolution. The proposed framework is applied to a smart city, which is one of the most interdisciplinary and fast-evolving areas. These findings indicate that the evolution of smart transportation and smart grids is likely to be the focus. The findings also indicate that newly emerging research may lack openness and diversity. This study contributes to further investigate research trends and planning research strategies for new and interdisciplinary areas.