This work copes with the problem of link prediction in large-scale two-mode social networks. Two variations tasks are studied: predicting links a bipartite graph and unimodal obtained by projection over one its node sets. For both tasks, we show an empirical way, that taking into account nature can enhance substantially performances models learn. is achieved introducing new topological atttributes to measure likelihood two nodes be connected. Our approach, for consists expressing as class...

The growing availability of multirelational data gives rise to an opportunity for novel characterization complex real-world relations, supporting the proliferation diverse network models such as Attributed Graphs, Heterogeneous Networks, Multilayer Temporal Location-aware Knowledge Probabilistic and many other task-driven data-driven models. In this paper, we propose overview these their main applications, described under common denomination Feature-rich i. e. where expressive power topology...

In this work we present a new approach for co-authorship link prediction based on leveraging information contained in general bibliographical multiplex networks. A network is graph defined over set of nodes linked by different types relations. For instance, the are studying here as follows : represent authors and links can be one following types: links, co-venue attending co-citing links. supervised-machine learning applied. formation model learned topological attributes describing both...

Multiplex network is an emergent model that has been lately proposed in order to cope with the complexity of real-world networks. A multiplex defined as a multi-layer interconnected graph. Each layer contains same set nodes but by different types links. This rich representation requires redefine most existing analysis algorithms. In this paper we focus on central problem community detection. Most approaches consist transforming problem, way or another, classical setting detection monoplex...

In this paper we propose a new topological approach for link prediction in dynamic complex networks. The proposed applies supervised rank aggregation method. This functions as follows: first the list of unlinked nodes network at instant t according to different measures (nodes characteristics aggregation, neighborhood based measures, distance etc). Each measure provides its own rank. Observing t+1 where some links appear, weight each performances predicting these observed links. These...

Leader-driven community detection algorithms (LdCD hereafter) constitute a new trend in devising for large-scale complex networks. The basic idea is to identify some particular nodes the target network, called leader nodes, around which local communities can be computed. Being based on computations, they are particularly attractive handle In this paper, we describe framework implementing LdCD algorithms, LICOD. We propose also way evaluating performances of algorithms. This consists...

Graph clustering techniques are very useful for detecting densely connected groups in large graphs. Many existing graph methods mainly focus on the topological structure, but ignore vertex properties. Existing have been recently extended to deal with nodes attribute. First we motivate interest study of this issue. Then review main approaches proposed problem. We propose a comparative some attributed network community detection algorithm both synthetic data and real world data.

In this work we tackle the problem of link prediction in co-authoring network. We apply a topological dyadic supervised machine learning approach for that purpose. A network is actually obtained by projection two-mode graph (an authoring linking authors to publications they have signed) over set. show performances can be substantially enhanced analyzing not only network, but also dual projecting original set publications.

In this work we propose a new efficient algorithm for communities construction based on the idea that community is animated by set of leaders are followed nodes. A node can follow different animating communities. The structured into two mains steps: identifying nodes in network playing role leaders, then assigning other to some leaders. We provide general framework implementing such an approach. First experimental results obtained applying real networks show effectiveness proposed

Link prediction is a central task in the field of dynamic complex network analysis. A major trend this area consists applying dyadic topological approach. Most existing approaches apply machine learning algorithms where link problem converted into binary classification task. In work, we propose new approach supervised social choice algorithm. Given training graph observed over period [t <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> , t...

Performance of cluster ensemble approaches is now known to be tightly related both quality and diversity input base clusterings. Cluster selection (CES) refers the process filtering raw set clusterings in order select a subset high diverse Most existing CES apply one index for measuring another evaluating Moreover number usually given as an function. In this work we propose new approach that allow taking into account indexes. addition, proposed computes automatically return. The basic idea...

In this paper we describe a new distributed collaborative bookmark system, called COWING (for COllaborative Web IndexING system). The system is composed of set assistant agents, WINGS, and central agent that manages the user's organization. Each user assisted by Wing performs two tasks: learning strategy in classifying her/his own bookmarks interacting with other WING agents order to fetch match local information need.

In this paper we propose a new approach for efficiently identifying ego-centered communities in complex networks. Most existing approaches are based on applying greedy optimisation process guided by given objective function. Different functions has been proposed the scientific literature, each capturing some specific feature of desired communities. work, to apply ensemble ranking order combine different functions. Preliminary Results obtained from experiments benchmark networks argue...

In this work, we explore applying a link prediction approach to tag recommendation in broad folksonomies. The original idea of the is mine dynamic tagging activity order compute most suitable for given user and resource. history each modeled by temporal sequence bipartite graphs linking tags resources. Given target resource, first set similar users. identified users merged one on graphs. obtained used learn model learned then applied predict be linked resource list top We get hence several...

In this work, we present an original seed-centric algorithm for community detection. Instead of expanding communities around selected seeds as most existing approaches do, propose applying ensemble clustering approach to different network partitions derived from local computed each seed. Local are themselves ranking that allow combining modularity functions used in a classical greedy optimization process.

In this paper, we target the problem of mining descriptive profiles computer network intrusion attacks. We present an exploratory and explanation-aware approach using subgroup discovery – facilitating human-in-the-loop interaction for guiding exploration process since results are inherently interpretable patterns. Furthermore, explore enriching feature set describing traffic (i. e., exchanged packets) with a new type features computed on complex networks depicting interactions among...