Link Analysis Algorithms in Hyperlinked Networks

Link analysis algorithms form the backbone of modern approaches to mining structure and influence in networks defined by hyperlinks. At their core lies the representation of a hyperlinked system as a directed graph, in which nodes correspond to pages or entities and edges to hyperlinks. Early methods such as eigenvector centrality and HITS introduced the concept of iteratively computing scores based on neighbours’ importance. The PageRank algorithm generalised this idea by modelling a random surfer who follows links with a probability of transition tempered by a damping factor, yielding a stationary distribution that ranks nodes by global prominence. Subsequent developments have explored dual perspectives on link flow: CheiRank inverts the PageRank process to emphasise outgoing links, while two-dimensional schemes such as 2DRank combine both PageRank and CheiRank vectors to capture nodes that are simultaneously authoritative and communicative. More recent work has introduced the reduced Google matrix, which condenses the full network into a smaller effective matrix that retains both direct and indirect interactions among a chosen subset of nodes. Collectively, these algorithms have found applications in web search, recommendation systems, social influence analysis, biological network inference and geopolitical studies, providing a principled, scalable framework for extracting patterns of connectivity, influence and community structure from vast hyperlinked datasets.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

A recent study in sports analytics has demonstrated an enhanced PageRank variant that integrates forward and backward propagation through a directed tournament network. By calibrating the algorithm on synthetic and real match results, it outperformed classical PageRank and simple win-count rankings in early tournament phases, and offered robust performance under varying levels of randomness introduced by factors such as home-team advantage. Another investigation into multilingual Wikipedia editions employed PageRank, CheiRank and 2DRank to rank articles in nine languages. This work revealed patterns of local versus global cultural prominence, identified ‘global heroes’ whose influence transcends linguistic communities, and mapped interconnections among topical fields, illustrating how link analysis exposes entangled knowledge structures across cultures. In the domain of biomedical knowledge, the reduced Google matrix approach has been applied to the network of Wikipedia articles on cancer types, therapies and countries. By distilling direct and indirect links among these entities, researchers constructed sensitivity networks that quantify how changes in one node propagate to others, yielding insights into global disease influence, drug interdependencies and regional vulnerabilities, and showcasing the power of condensed link-analysis matrices for targeted network interrogation.

Link Analysis Algorithms in Hyperlinked Networks publication trend

The graph below shows the total number of articles in link analysis algorithms in hyperlinked networks across all publications each year (not limited to Nature Index journals).

Technical terms

Hyperlinked network: A directed graph in which nodes represent documents or entities and edges correspond to hyperlinks, encoding the navigational structure of a dataset.

PageRank: A stochastic link-analysis algorithm that assigns node importance based on the stationary distribution of a random walker following hyperlinks with a damping probability.

Google matrix: The transition probability matrix combining the hyperlink adjacency matrix with a teleportation factor, used to compute PageRank scores.

Reduced Google matrix: A smaller matrix derived from the full Google matrix that preserves direct and indirect interactions among a selected subset of nodes for focused analysis.

CheiRank: A variant of PageRank computed on the network with inverted link directions, highlighting nodes with high communicative or outgoing influence.

2DRank: A bi-dimensional ranking scheme that jointly considers PageRank and CheiRank scores to identify nodes that are both authoritative and influential communicators.

References

1. Improving PageRank using sports results modeling. Knowledge-Based Systems (2022).
2. Highlighting Entanglement of Cultures via Ranking of Multilingual Wikipedia Articles. PLOS ONE (2013).
3. Wikipedia network analysis of cancer interactions and world influence. PLOS ONE (2019).