Skip to Main content Skip to Navigation
Theses

SAFE-NEXT : UNE APPROCHE SYSTEMIQUE POUR L'EXTRACTION DE CONNAISSANCES DE DONNEES.
Application A La Construction Et A L'interprétation De Scénarios D'accidents De La Route

Abstract : Nowadays, given the automation of data collection, very large databases are constructed. The exploitation of these data in accidentology and several others fields (e.g. marketing, engineering, etc.) requires automatic techniques of Knowledge Discovery in Databases (KDD). Incorporating expert knowledge in the KDD process is fundamental to handle the complexity of data, domain and knowledge. This necessitates the development of approaches, methods and techniques intended to identify, represent and operationalize expert knowledge.
In this dissertation, we propose a new approach, SAFE-Next (Systemic Approach For Enhanced kNowledge EXTraction), which integrates the following four approaches: the first one, ASMEC (Approche Systémique de ModElisation des Connaissances), allows knowledge modeling according to multiple viewpoints and granularity levels. The second approach, AICEF (Approche d'Incorporation des Connaissances Expertes dans la Fouille de données), uses the ASMEC knowledge model to elaborate multi-view metadata. It then uses these metadata as a tool for incorporating expert knowledge into the KDD process. The third approach, ASAIC (Approche Systémique d'Analyse d'Impact de Changement), uses the ASMEC knowledge model to carry out multi-view change impact analysis. The fourth approach, ASEM (Approche Systémique d'Evaluation de Modèles), provides an assessment framework for knowledge models.
The epistemological and methodological foundations of our work are constructivism and systemic approach (or cybernetics). Based on these backgrounds, our research contributions concern several disciplines, ranging from Accidentology, Knowledge Engineering, Knowledge Discovery in Databases and Design. In accidentology, SAFE-Next provides experts with an efficient tool for knowledge management. It enables the elaboration of multi-view accident scenarios, which are a powerful tool for understanding accident mechanisms in order to develop safety counter-measures. Furthermore, SAFE-Next provides a knowledge capitalization tool. In knowledge engineering, SAFE-Next supplies, via ASMEC, a multi-view knowledge model and thereby allows the integration of different viewpoints stemming from different users. Furthermore, it provides a multi-granularity knowledge model and in that way addresses the difficulty of knowledge identification and formalization. At the same time, SAFE-Next permits, via ASEM, the evaluation of knowledge models, an issue rarely addressed in literature. In Knowledge Discovery in Database, SAFE-Next enables, via AICEF, the incorporation of domain knowledge in the data preprocessing step (i.e. the first step in a KDD process) and more specifically in the attribute selection task. Likewise, the multi-view metadata enables the incorporation of domain knowledge in the interpretation step (i.e. the last step in a KDD process). In design, SAFE-Next provides safety system developers with an efficient tool to construct the design space. Scenarios enable them to understand complex behaviors and thereby to define solutions and alternatives. SAFE-Next also provides, via ASAIC, an approach for multi-view change impact analysis. Moreover, it proposes an extension of the change impact analysis to the use process of a given product as well as the evaluation process instead of limiting it to the design process.
Document type :
Theses
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00011540
Contributor : Walid Ben Ahmed <>
Submitted on : Wednesday, February 8, 2006 - 5:20:05 PM
Last modification on : Monday, October 19, 2020 - 11:08:38 AM
Long-term archiving on: : Monday, September 20, 2010 - 2:14:35 PM

Identifiers

  • HAL Id : tel-00011540, version 2

Collections

Citation

Walid Ben Ahmed. SAFE-NEXT : UNE APPROCHE SYSTEMIQUE POUR L'EXTRACTION DE CONNAISSANCES DE DONNEES.
Application A La Construction Et A L'interprétation De Scénarios D'accidents De La Route. domain_stic.gest. Ecole Centrale Paris, 2005. Français. ⟨tel-00011540v2⟩

Share

Metrics

Record views

562

Files downloads

2055