XAI : a comparative study of post-hoc and intrinsic explainability methods for categorical tabular data

Godfrin, Siméon
(2026)

Files

GODFRIN_60921800_2026_metrics.zip
  • Closed access
  • Unknown
  • 49.55 KB
GODFRIN_60921800_2026_NeuralN.zip
  • Closed access
  • Unknown
  • 5.07 KB
GODFRIN_60921800_2026_preprocessing.zip
  • Closed access
  • Unknown
  • 216 B
GODFRIN_60921800_2026.pdf
  • Closed access
  • Adobe PDF
  • 3.15 MB
GODFRIN_60921800_2026_conda_env.yml
  • Closed access
  • Unknown
  • 12.62 KB

Details

Supervisors
Caelen, Olivier ; Vande Kerckhove, Corentin ; Van Oirbeek, Robin
Faculty
Faculté des sciences
Degree label
Master [120] en science des données, orientation statistique, à finalité spécialisée
Abstract
enThis thesis compared LIME, KernelSHAP, DiCE, and an intrinsic interpretable model OSDT (Optimal Sparse Decision Tree) under a functionally grounded benchmark ( in a multi-class classification setting on an almost fully categorical tabu- lar dataset.