Skip to Main content Skip to Navigation
Theses

Apprentissage numérique et symbolique pour le diagnostic et la réparation automobile

Tom Obry 1
1 LAAS-DISCO - Équipe DIagnostic, Supervision et COnduite
LAAS - Laboratoire d'analyse et d'architecture des systèmes
Abstract : Clustering is one of the methods resulting from unsupervised learning which aims to partition a data set into different homogeneous groups in the sense of a similarity criterion. The data in each group then share common characteristics. DyClee is a classifier that performs a classification based on digital data arriving in a continuous flow and which proposes an adaptation mechanism to update this classification, thus performing dynamic clustering in accordance with the evolution of the system or process being followed. Nevertheless, the only consideration of numerical attributes does not allow to apprehend all the fields of application. In this generalization objective, this thesis proposes on the one hand an extension to nominal categorical data, and on the other hand an extension to mixed data. Hierarchical clustering approaches are also proposed in order to assist the experts in the interpretation of the obtained clusters and in the validation of the generated partitions. The presented algorithm, called Mixed DyClee, can be applied in various application domains. In the case of this thesis, it is used in the field of automotive diagnostics.
Document type :
Theses
Complete list of metadata

https://hal.laas.fr/tel-02961151
Contributor : Abes Star :  Contact
Submitted on : Tuesday, March 30, 2021 - 12:10:14 PM
Last modification on : Thursday, June 10, 2021 - 3:05:47 AM

File

2020TomOBRY.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02961151, version 2

Citation

Tom Obry. Apprentissage numérique et symbolique pour le diagnostic et la réparation automobile. Automatique / Robotique. INSA de Toulouse, 2020. Français. ⟨NNT : 2020ISAT0014⟩. ⟨tel-02961151v2⟩

Share

Metrics

Record views

87

Files downloads

29