Interpretability assessment of fuzzy knowledge bases

摘要: Computing with words (CWW) relies on linguistic representation of knowledge that is processed by operating at the semantical level defined through fuzzy sets. Linguistic a major issue when rule based models are acquired from data some form empirical learning. Indeed, these often requested to exhibit interpretability, which normally evaluated in terms structural features, such as complexity, properties sets and partitions. In this paper we propose different approach for evaluating interpretability notion cointension. The rule-based model measured cointension degree between explicit semantics, formal parameter settings model, implicit semantics conveyed reader knowledge. Implicit calls user's difficult externalise. Nevertheless, identify set -- call “logical view” expected hold used our evaluate semantics. practice, new base obtained minimising logical properties. Semantic comparison made performances two bases, supposed be similar almost equivalent. If case, deduce view applicable can tagged interpretable viewpoint. These ideas then define strategy assessing classifiers (FRBCs). has been pre-existent FRBCs, learning processes well-known benchmark dataset. Our analysis highlighted them not cointensive knowledge, hence their appropriate, even though they point view.