logolpt fri_logo

Network and link analysis

Network analysis strives towards revealing the structure of large and complex real-world networks. These include (online) social networks, web graphs, Internet maps, software system networks, power grids, road networks, biological and ecological networks, and other. Link analysis can be seen as data mining over relational data represented as networks. Main challenges are thus classification, clustering and ranking of network nodes in the context of their imminent neighbors, while special emphasis is put on the applicability of the developed techniques.

Network analysis

Our work on network analysis is mainly focused on different mesoscopic structures in real-world networks including communities and functional modules, and on common macroscopic properties of networks including clustering, self-similarity, structural-worlds and other. We have proposed different algorithms for module detection and various network models, identified several common properties of real-world networks and analyzed software networks in greater detail. For more see lovro.lpt.fri.uni-lj.si.

Structural modules
Structural modules
Network clustering assortativity
Network clustering assortativity
Self-similar scaling of density
Self-similar scaling of density
Community detection algorithms
Community detection algorithms

Link analysis

Our work on link analysis in the past was mostly focused on classification and ranking of nodes in networks of traffic accidents, as a part of an expert system for identifying collaborating groups of motor insurance fraudsters. For more see lovro.lpt.fri.uni-lj.si.

Traffic accidents networks
Traffic accidents networks
Iterative assessment algorithm
Iterative assessment algorithm
Motor insurance fraudsters
Motor insurance fraudsters

Work on link analysis is directly related with the field of information extraction, where similar techniques are used for, e.g., coreference resolution.