Link Prediction in Networks with Core-Fringe Data

Abstract

Data collection often involves the partial measurement of a larger system. A common example arises in collecting network data: we often obtain network datasets by recording all of the interactions among a small set of core nodes, so that we end up with a measurement of the network consisting of these core nodes along with a potentially much larger set of fringe nodes that have links to the core. Given the ubiquity of this process for assembling network data, it is crucial to understand the role of such a "core-fringe" structure.Here we study how the inclusion of fringe nodes affects the standard task of network link prediction. One might initially think the inclusion of any additional data is useful, and hence that it should be beneficial to include all fringe nodes that are available. However, we find that this is not true; in fact, there is substantial variability in the value of the fringe nodes for prediction. Once an algorithm is selected, in some datasets, including any additional data from the fringe can actually hurt prediction performance; in other datasets, including some amount of fringe information is useful before prediction performance saturates or even declines; and in further cases, including the entire fringe leads to the best performance. While such variety might seem surprising, we show that these behaviors are exhibited by simple random graph models.1 This distinction between core and fringe is fundamentally driven by measurement of the available data; we have measured all interactions involving members of the core, and this brings the fringe indirectly into the data. As such, it is distinct from work on the core-periphery structure of networks, which typically refers to settings in which the core and periphery both fully belong to the measured network, and the distinction is in the level of centrality or status that the core has relative to the periphery [7,16,31,37].

Fast indexing algorithm for efficient k NN queries on complex networks

“…By contrast, a tree is a long stretched sparse subgraph. As reported in [11], the vast majority of nodes are included in trees, and only 0.6 -9.3% nodes are contained in the core for a typical complex network.…”

“…In lines 8-9, the algorithm constructs a set of nodes V c , which includes all non-tree nodes and all root nodes in T. Algorithm 1 then links nodes in V c if they have edges in E (line 10). Finally, the algorithm updates the weights of edges (line [11][12]. As shown in line 12, each weight is set as the shortest path distance between two adjacent core nodes.…”

Community Detection Based on the Nodes Role in a Network: The Telegram Platform Case

“…The problem we face here is that Telegram does not provide researchers with information of the readers' level. This data is useful for structure analysis in the case of social media [1].…”

“…To define communities in a network, we need to find similar or close nodes. Benson et al (2019) as a similarity evaluator use the number of common neighbors and the Jaccard similarity of the neighbor sets [1]. We, based on the idea of strategy, use seven indicators.…”

Stochastic Opinion Dynamics for Interest Prediction in Social Networks

“…Moreover, another key reason for which we choose to work with this limited information is that the network may be huge, thus it may be prohibiting to mine all -or at least a large fraction -of its edges. In fact, this is a serious practical consideration, taking into account the massive growth of modern OSN, and a considerable volume of work [5,40,55] has devoted to addressing it. The influencers' sublinear number allows for a quite fast initialization (in worst-case strongly subquadratictime) of the peripheral users' interests.…”