Interpretable Regional Descriptors: Hyperbox-Based Local Explanations

Dandl, Susanne ; Casalicchio, Giuseppe ; Bischl, Bernd und Bothmann, Ludwig (2023): Interpretable Regional Descriptors: Hyperbox-Based Local Explanations. ECML PKDD 2023, Torino, Italy, September 18 -22 2023. Koutra, Danai; Plant, Claudia; Gomez Rodriguez, Manuel; Baralis, Elena und Bonchi, Francesco (eds.) : In: Machine Learning and Knowledge Discovery in Databases : Research Track : European conference, ECML PKDD 2023, Turin, Italy, September 18-22, 2023 : proceedings, Vol. 14171 Cham: Springer. pp. 479-495

Abstract

This work introduces interpretable regional descriptors, or IRDs, for local, model-agnostic interpretations. IRDs are hyperboxes that describe how an observation’s feature values can be changed without affecting its prediction. They justify a prediction by providing a set of “even if” arguments (semi-factual explanations), and they indicate which features affect a prediction and whether pointwise biases or implausibilities exist. A concrete use case shows that this is valuable for both machine learning modelers and persons subject to a decision. We formalize the search for IRDs as an optimization problem and introduce a unifying framework for computing IRDs that covers desiderata, initialization techniques, and a post-processing method. We show how existing hyperbox methods can be adapted to fit into this unified framework. A benchmark study compares the methods based on several quality measures and identifies two strategies to improve IRDs.

• Item Type: Conference or Workshop Item (Paper)
• Faculties: Mathematics, Computer Science and Statistics > Statistics
• Subjects: 000 Computer science, information and general works > 004 Data processing computer science; 500 Science > 510 Mathematics
• ISBN: 978-3-031-43417-4 ; 978-3-031-43418-1
• Place of Publication: Cham
• Language: English