Etienne Cuvelier

A buzz and e-reputation monitoring tool for twitter based on galois lattices

In the actual interconnected world, the speed of broadcasting of information leads the formation of opinions towards more and more immediacy. Big social networks, by allowing distribution, and therefore broadcasting of information in a almost instantaneous way, also speed up the formation of opinions concerning actuality. Then, these networks are great observatories of opinions and e-reputation. In this e-reputation monitoring task, it is easy to get a set of information (web pages, blog pages, tweets,...) containing a chosen word or a set of words ( a company name, a domain of interest,...), and then we can easily search for the most used words. But a harder, but more interesting task, is to track the set of jointly used words in this dataset, because this latter contains the more shared advice about the initial searched set of words. Precisely, the exhaustive discovering of the shared properties of a collection of objects is the main task of the Galois lattices used in the Formal Concept Analysis. In this article we state clearly the characteristics, advantages and constraints of one of the more successful online social networks: Twitter. Then we detail the difficult task of tracking, on Twitter, the most forwarded information about a chosen subject. We also explain how the characteristics of Galois lattices permit to solve elegantly and efficiently this problem. But, retrieving the most used corpus of words is not enough, we have to show the results in an informative and readable manner, which is not easy when the result is a Galois Lattice. Then we propose a visualisation called topigraphic network of tags, which represent a tag cloud in a network of concepts with a topographic allegory, which permits to visualise the more important concepts found about a given search on Twitter.

Graph Mining and Communities Detection

The incredible rising of on-line social networks gives a new and very strong interest to the set of techniques developed since several decades to mining graphs and social networks. In particular, community detection methods can bring very valuable informations about the structure of an existing social network in the Business Intelligence framework. In this chapter we give a large view, firstly of what could be a community in a social network, and then we list the most popular techniques to detect such communities. Some of these techniques were particularly developed in the SNA context, while other are adaptations of classical clustering techniques. We have sorted them in following an increasing complexity order, because with very big graphs the complexity can be decisive for the choice of an algorithm.

A Probability Distribution Of Functional Random Variable With A Functional Data Analysis Application

Probability distributions are central tools for probabilistic modeling in data mining, and they lack in functional data analysis (FDA). In this paper we propose a probability distribution law for functional data. We build it using jointly the quasi-arithmetic means and the generators of Archimedean copulas. We also define a density adapted to the infinite dimension of the space of functional data. For this we use the Gateaux differential. We illustrate the utility of this tool in FDA, applying it in a mixture decomposition classification

DB2SNA: An All-in-One Tool for Extraction and Aggregation of Underlying Social Networks from Relational Databases

In the enterprise context, People need to visualize different types of interactions between heterogeneous objects (e.g. product and site, customers and product, people interaction (social network)…). The existing approaches focus on social networks extraction using web document. However a considerable amount of information is stored in relational databases. Therefore, relational databases can be seen as rich sources for extracting a social network. The extracted network has in general a huge size which makes it difficult to analyze and visualize. An aggregation step is needed in order to have more understandable graphs. In this chapter, we propose a heterogeneous object graph extraction approach from a relational database and we present its application to extract social network. This step is followed by an aggregation step in order to improve the visualisation and the analyse of the extracted social network. Then, we aggregate the resulting network using the k-SNAP algorithm which produces a summarized graph.

Soft and Adaptive Aggregation of Heterogeneous Graphs with Heterogeneous Attributes

In the enterprise context, people need to exploit, interpret and mainly visualize different types of interactions between heterogeneous objects. Graph model is an appropriate way to represent those interactions. Nodes represent the individuals or objects and edges represent the relationships between them. However, extracted graphs are in general heterogeneous and large sized which makes it difficult to visualize and to analyze easily. An adaptive aggregation operation is needed to have more understandable graphs in order to allow users discovering underlying information and hidden relationships between objects. Existing graph summarization approaches such as k-SNAP are carried out in homogeneous graphs where nodes are described by the same list of attributes that represent only one community. The aim of this work is to propose a general tool for graph aggregation which addresses both homogeneous and heterogeneous graphs. To do that, we develop a new soft and adaptive approach to aggregate heterogeneous graphs i.e., composed of different node attributes and different relationship types using the definition of Rough Set Theory RST combined with Formal Concept Analysis FCA, the well known K-Medoids and the hierarchical clustering methods. Aggregated graphs are produced according to user-selected node attributes and relationships. To evaluate the quality of the obtained summaries, we propose two quality measures that evaluate respectively the similarity and the separability in groups based on the notion of common neighbor nodes. Experimental results demonstrate that our approach is effective for its ability to produce a high quality solution with relevant interpretations.

An approach to Stochastic Process using Quasi-Arithmetic Means

Probability distributions are central tools for probabilistic modeling in data mining. In functional data analysis (FDA) they are weakly studied in the general case. In this paper we discuss a probability distribution law for functional data considered as stochastic process. We define first a new kind of stationarity linked t o the Archimedean copulas, and then we build a probability distribution using jointly the Quasi-arithmetic means and the generators of Archimedean copulas. We also study some properties of this new mathematical tool.

Classification de données fonctionnelles par décomposition de mélange Apports de la visualisation dans le cas des distributions de probabilité

Functional data can come from repeated measures, but also as result of statistical analysis. In symbolic data analysis a symbolic object can be described with a probability distribution. The clustering of such objects can be performed using a mixture decomposition with archimedean copulas on values of the distributions computed in q points, named intersection points. So far this points were chosen randomly. In this paper, using visualizations, we try, empirically, to understand what is the best choice for the number and the location of these intersections points. We propose also some rules to choose this parameter of the classification.

Parametric Families of Probability Distributions for Functional Data Using Quasi-Arithmetic Means with Archimedean Generators

Parametric probability distributions are central tools for probabilistic modeling in data mining, and they lack in functional data analysis (FDA). In this paper we propose to build this kind of distribution using jointly Quasi-arithmetic means and generators of Archimedean copulas. We also define a density adapted to the infinite dimension of the space of functional data. We use these concepts in supervised classification. 1. QAMML distributions Let (Ω,A, P ) a probability space and D a closed real interval. A functional random variable (frv) is any function from D×Ω → R such for any t ∈ D, X(t, .) is a real random variable on (Ω,A, P ). Let L(D) be the space of square integrable functions (with respect to Lebesgues measure) u(t) defined on D. If f, g ∈ L(D), then the pointwise order between f and g on D is defined as follows : ∀t ∈ D, f(t) ≤ g(t) ⇐⇒ f ≤D g. (1) It is easy to see that the pointwise order is a partial order over L(D), and not a total order. We define the functional cumulative distribution function (fcdf) of a frv X on L(D) computed at u ∈ L(D) by : FX,D(u) = P [X ≤D u]. (2) To compute the above probability, let us remark that, it is easy to compute the probability distribution of the value of X(t) for a specific value of t, and this for any t ∈ D. Then we define respectively the surface of distributions and the surface of densities as follow : G : D × R → [0, 1] : (t, y) 7→ P [X(t) ≤ y] (3) g : D × R → [0, 1] : (t, y) 7→ ∂ ∂t G (t, y) (4) We can use various methods for determining suitable g and G for a chosen value of X. Thus for example, if X is a Gaussian process with mean value μ(t) and standard deviation σ(t), then, for any (t, y) ∈ D × R, we have : G (t, y) = FN (μ(t),σ(t))(y) and g (t, y) = fN (μ(t),σ(t))(y). In the following we will always use the function G with a function u of L (D), so, for the ease of the notations, we will write : G [t;u] = G [t, u (t)]. We will use the same notation for g. In what follows we define our parametric families of probability distributions. Let X be a frv, u ∈ L(D) and G its Surface of Distributions. Let also φ be a continuous strictly decreasing function from [0, 1] to [0,∞] such that φ(0) = ∞, φ(1) = 0, where ψ = φ must be completely monotonic on [0,∞[ i.e. (−1) d k dtk ψ(t) ≥ 0 for all t in [0,∞[ and for all k. We define the Quasi-Arithmetic Mean of Margins Limit (QAMML) distribution of X by : FX,D(u) = ψ [

Symbolic Markov Chains

Stochastic processes have, since a long time, large applications in quite different domains. The standard theory considers discrete or continuous state space. We consider here the concept of Stochastic Process associated to all the cases of symbolic variables: quantitative, categorical single and multiple, interval, modal. More particularly, we adapt the definition of Markov Chain and give the equivalent of the Chapman-Kolmogorov theorem in all cases.