Fuzzy classifiers with information granules in feature space and logic-based computing

Highlights：

• we carefully analyzed the concept of fuzzy classifiers elaborating on their functional modules and their design process.

• we present a thorough generalization of fuzzy classifiers by involving various t-norms and t-conorms and compare the performance of different construct strategies.

• we propose an augmentation method for the fuzzy classifier by introducing an interaction to quantify the strength of connection between the fuzzy rules and membership.

• we contrast the quality of fuzzy classifiers with those commonly non-fuzzy classifiers and identify situations where fuzzy sets used in classifiers outperforms their counterparts.

摘要

•we carefully analyzed the concept of fuzzy classifiers elaborating on their functional modules and their design process.•we present a thorough generalization of fuzzy classifiers by involving various t-norms and t-conorms and compare the performance of different construct strategies.•we propose an augmentation method for the fuzzy classifier by introducing an interaction to quantify the strength of connection between the fuzzy rules and membership.•we contrast the quality of fuzzy classifiers with those commonly non-fuzzy classifiers and identify situations where fuzzy sets used in classifiers outperforms their counterparts.

论文关键词：Fuzzy classifiers,Performance analysis,Logic processing,Receiver operating characteristics (ROC),Triangular norms,Fuzzy clustering,Particle swarm optimizer (PSO)

论文评审过程：Received 21 September 2017, Revised 16 January 2018, Accepted 11 March 2018, Available online 12 March 2018, Version of Record 22 March 2018.