Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/84720
標題: Data-Driven Interval Type-2 Neural FuzzySystem With High Learning Accuracy andImproved Model Interpretability
關鍵字: Fuzzy neural networks (FNNs), interpretable fuzzy systems (FSs);sequence prediction;type-2 FSs
Project: IEEE TRANSACTIONS ON CYBERNETICS, Volume 43, Issue 6.
摘要: 
Current studies of type-2 neural fuzzy systems (FSs)(NFSs) primarily focus on building a fuzzy model with highaccuracy and disregard the interpretability of fuzzy rules. Thispaper proposes a data-driven interval type-2 (IT2) NFS withimproved model interpretability (DIT2NFS-IP). The DIT2NFS-IPuses IT2 fuzzy sets in its antecedent part and intervals in itszero-order Takagi–Sugeno–Kang-type consequent part for ruleform simplicity. The initial rule base is generated by a self-splittingclustering algorithm in the input–output space. The DIT2NFS-IPuses a two-phase parameter-learning algorithm to design an accuratemodel with improved rule interpretability. In the firstphase, a new cost function that considers both accuracy andtransparent fuzzy set partition is defined. The antecedentand consequent parameters are learned through gradient descentand rule-ordered recursive least squares algorithms, respectively,to achieve cost function minimization. The second phase performsa fuzzy set reduction, followed by consequent parameter learningto improve accuracy. Comparisons with different type-1 andtype-2 FSs in five databased modeling and prediction problemsverify the performance of the DIT2NFS-IP in both model accuracyand interpretability.
URI: http://hdl.handle.net/11455/84720
Appears in Collections:電機工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.