Informetric Measures In Weighted Information Networks

As one of the four major types in network science, information network refers to a set of information entities (nodes) and their information behaviors (links). Most real world information networks have weighted links. And the link weights usually represent significant meaning for measurement. However, seldom current basic quantitative methods, mainly from social network analysis in sociology, grasp such structural characteristics of information networks accurately.According to the inspiration of informetrics, this paper attempts to give the measures of weighted information network, from the local nodes and links to the whole network.(1) First, for nodes, we give the method of h-degree, and focus on the link weight of nodes and the structural information. Based on h-degree, we propose h-centrality、 h-centralization、h-ratio and other relevant methods, and apply the method in directed networks. In addition to the theoretical discussion of the nature in these measures, several networks of publications, journals and research fields empirically show the characteristics and validity of h-degree.(2) For links, we investigate whether the link strengths follow the power-law distribution in weighted networks. After testing 12 different paper co-citation networks with two methods, fitting in double-log scales and the Kolmogorov-Smirnov test (K-S test), we observe that the link strengths also follow the approximate power-law distribution in most cases. The results suggest that the power-law type distribution could emerge not only in nodes and informational entities, but also in links and informational connections. Based on this finding, we give an option for exploring links in networks, called h-strength, with explicit focuses on the insight of links and their strengths. The h-strength and its extensions can naturally simplify a complex network to a small and concise sub-network (h-subnet) but retains the most important links with its core structure. Its applications in two typical information networks, the paper co-citation network of the topic ’h-index’ and five scientific collaboration networks in the field’water recourses’, suggest that h-strength and its extensions could be a useful choice for simplifying and visualizing links in complex network. Moreover, it is observed that the two informetric models, Glanzel-Schubert model and Hirsch model, roughly hold in the context of h-strength.(3) For whole works, a simple method, denoted as h-bone, is introduced for abstracting the network mainstay. This method includes three steps:(a) mining the core nodes; (b) extracting the core links; (c) resetting the’bridge’. h-bone involves structures of weights and betweenness. Thus it could be a comprehensive framework. The analysis of a large scale paper co-citation network shows that, h-bone successfully simplify the complex network to a concise bone. One advantage of the method is h-bone just use few links and nodes but retain the most important components of the case network.Measurement occurs as the infrastructure work of network research. The measures given by this work is expected to constitute an optional method in the future information network research. Beginning with the method meeting the characteristics of information network more, it is expected to deeply understanding the real structures of information world.