Interpetability improvements to find the balance interpretability-accuracy in fuzzy modeling: An overview (Q4437848)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 2012425
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Interpetability improvements to find the balance interpretability-accuracy in fuzzy modeling: An overview |
scientific article; zbMATH DE number 2012425 |
Statements
2003
0 references
fuzzy modeling
0 references
interpretability
0 references
accuracy
0 references
rule-based systems
0 references
reduction
0 references
aggregation
0 references
local regression models
0 references
selection of input variables
0 references
simplification
0 references
exception rules
0 references
orthogonal transformations
0 references
0.8161007761955261
0 references
0.8091378808021545
0 references
0.8072604537010193
0 references
0.7987111210823059
0 references
0.778644859790802
0 references
Interpetability improvements to find the balance interpretability-accuracy in fuzzy modeling: An overview (English)
0 references
The study addresses an important issue of a tradeoff between interpretability and accuracy in fuzzy modeling. First it presents a general taxonomy of fuzzy models distinguishing between linguistic and precise fuzzy modeling, highlighting various tradeoffs occurring there. The taxonomy is cast in the framework of rule-based systems with a natural categorization involving linguistic rules (Mamdani-type) and local regression models (developed in the Takagi-Sugeno framework). In the case of the framework of the Mamdani-type of modeling, a systematic analysis of interpretability focuses on several underlying approaches such as selection of input variables and determination of input variables occurring in individual rules. The reduction or simplification of rule-bases deals with (a) merging/selecting subsets of rules, (b) forming disjunctive (DNF) forms of rules in which several linguistic terms defined in the same space are aggregated, and (c) building exception rules. In the context of interpretability of precise fuzzy models, the main trends concern various forms of ordering/selecting fuzzy rules, which involve orthogonal transformations (orthogonal least-squares methods, OLS, and singular value decomposition, SVD-QR). The enhanced interpretability of the Takagi-Sugeno rules is achieved by controlling the character of the conclusion part, viz. ``local'' models (obtained, e.g., through admitting a smooth polynomial of higher order). For the entire collection see [Zbl 1015.00012].
0 references