Predicting of links and weights together in complex networks: Prediction of citation count of scientists

Interactions and collaborations between entities in many different areas, especially social networks in the Internet, create more complex networks. In recent times, network analysis and data mining in complex networks have attracted the attention of researchers. Link prediction in complex networks is also one of the most interesting research topics. In this study, two methods were proposed for link prediction in complex networks. The aim of the first method is to predict citation count of scientists. In the proposed method, predicting citation count of scientists problem has been formulated as a link prediction problem in citation networks. A temporal link prediction metric has been proposed that takes into account upward/downward trends throughout the evolution of citation networks over time. The proposed link prediction approach is the first study that predicts links with its weights in the directed, weighted, and temporal networks. The experimental results on citation networks show the accuracy of the proposed method to predict citation count of scientists. The proposed link prediction metric was also compared with the classical link prediction metric, and it has been shown that the proposed measure is an effective link prediction metric in the test results. The aim of the second method is to increase the accuracy of neighborhood-based link prediction metrics by considering the role of link direction information in link formation in the directed networks. The role of link direction information in link formation has not been considered in the majority of link prediction metrics in the literature. For this purpose, in this study, a general method is proposed in which the classical neighborhood based link prediction metrics are calculated by using on directional network motifs. Test results on directed networks have shown that the proposed method considerably improves the accuracy of neighborhood-based link prediction metrics.