Skip to Main content Skip to Navigation

Représentations visuelles adaptatives de connaissances associant projection multidimensionnelle (MDS) et analyse de concepts formels (FCA)

Abstract : Information retrieval tools are faced with the constant increase of data both in volume and in dimensionality and the traditional list of results no longer meet many applications' requirements. New visual representation techniques are needed. These new techniques have to provide an overview of large and multidimensional data sets that gives insights into the underlying trends and structures. They must also be able to represent, in detail, portions of the original data from different standpoints. The aim is to assist the user in her data exploration task by designing a shrewd link between general and local views, that maintains her mental map. In order to achieve this goal, we develop a combination of data analysis techniques that identify pertinent portions of data as well as information visualization techniques that intuitively and dynamically explore these portions of data in detail. In addition, a formalization of the visualization process is needed. We introduce a formal frame that is used to specify visualizations from data structures. Concretely, the solution proposed is an original navigation method that combines techniques from Formal Concept Analysis (FCA) and Multi-Dimensional Scaling (MDS) visualization approaches to suggest navigation paths in the data. This method is based on the "overview + detail" paradigm: One component is an overall view which summarises the underlying structure of the data. A second component is a local view showing an element of the overall view in detail. We take advantage of the classification skills of the Galois lattice by using it as the overall view that reveals the inner data structure and suggests possible navigation paths. The local view uses Multi-Dimensional Scaling to display the objects in the extent of a selected concept. We illustrate and discuss the pertinence of our method on concrete data sets, provided by our industrial partners, and show how hybridisation of FCA and traditional data visualization approaches, which have sometimes been considered distinct or incompatible, can be complementary.
Document type :
Domain :
Complete list of metadata

Cited literature [90 references]  Display  Hide  Download
Contributor : Ecole Mines ParisTech Connect in order to contact the contributor
Submitted on : Monday, August 3, 2009 - 8:00:00 AM
Last modification on : Tuesday, March 8, 2022 - 4:14:02 PM
Long-term archiving on: : Saturday, November 26, 2016 - 6:40:10 PM


  • HAL Id : pastel-00004559, version 1


Jean Villerd. Représentations visuelles adaptatives de connaissances associant projection multidimensionnelle (MDS) et analyse de concepts formels (FCA). domain_stic. École Nationale Supérieure des Mines de Paris, 2008. Français. ⟨NNT : 2008ENMP1622⟩. ⟨pastel-00004559⟩



Record views


Files downloads