Reference : Robustness analysis of clustering and classification techniques
Dissertations and theses : Doctoral thesis
Physical, chemical, mathematical & earth Sciences : Mathematics
Robustness analysis of clustering and classification techniques
Ruwet, Christel mailto [Université de Liège - ULiège > Département de mathématique > Statistique mathématique >]
Université de Liège, ​​Belgique
Docteur en Sciences
Haesbroeck, Gentiane mailto
Garcia-Escudero, Luis Angel mailto
Albert, Adelin mailto
Heuchenne, Cédric mailto
Croux, Christophe mailto
Dehon, Catherine mailto
Gordaliza, Alfonso mailto
[en] Clustering ; Classification ; Error rate ; Breakdown point ; Influence function ; Robustness
[en] As mentioned in the title, the framework of this doctoral dissertation encompasses two different subjects: robust statistics on the one hand and classification and clustering techniques on the other hand.
Robust procedures try at the same time to emulate classical procedures and to produce results that are not unduly affected by contaminated observations or deviations from model assumptions.
Classification and clustering techniques try to find groups among observations. Grouping is one of the most basic abilities of living creatures; the simple fact of naming objects is already grouping. The main interest lies in the fact that the characteristics of a group as well as its differences from other groups can be used as a summary of the dataset.

File(s) associated to this reference

Fulltext file(s):

Restricted access
These.pdfAuthor preprint2.52 MBRequest copy

Additional material(s):

File Commentary Size Access
Restricted access
DefenseBis.pdfSlides de la défense1.04 MBRequest copy
Restricted access
Erratum.pdfErratum31.26 kBRequest copy

Bookmark and Share SFX Query

All documents in ORBi are protected by a user license.