Interpretability of Fuzzy Temporal Models
The paper presents a new approach to assessment of interpretability of fuzzy models. The approach differs from conventional ones, which consider interpretability from the point of structural complexity of both fuzzy model and its elements. In terms of developed approach, the interpretability means the ability of fuzzy model to reflect the same information presented in different forms to different users. Different forms of fuzzy model are given by use of specific inference system, which provides equivalent transformations of fuzzy rules from knowledge base on the linguistic level.
In our work, the inference system providing the equivalent transformations of fuzzy rules is developed for the specific class of fuzzy-temporal models. The necessary and sufficient conditions for properties of fuzzy rules are found. Such conditions provide semantic equivalence for equations obtained during fuzzy inference.
The formalized criterion is presented for interpretability of fuzzy model. The criterion is based on ability of model to keep information semantics on the fuzzy sets level when it is changed on the linguistic level.
KeywordsAssesment of fuzzy models Cointension Fuzzy interpretation of subjective information
The work was supported by RFBR (Grants No. 17-20-01040 ofi_m_RZD, No. 16-07-00032-a and No. 16-07-00086-a).
- 7.Alonso, J.M., Magdalena, L.: Combining user’s preferences and quality criteria into a new index for guiding the design of fuzzy systems with a good interpretability-accuracy trade-off. In: 2010 IEEE International Conference on Fuzzy Systems (FUZZ), pp. 961–968 (2010)Google Scholar
- 11.Marquez, A.A., Marquez, F.A., Peregrin, A.: A multi-objective evolutionary algorithm with an interpretability improvement mechanism for linguistic fuzzy systems with adaptive defuzzification. In: IEEE International Conference on Fuzzy Systems (FUZZ), pp. 277–283 (2010)Google Scholar
- 13.Riid, A., Rustern, E.: Interpretability improvement of fuzzy systems: reducing the number of unique singletons in zeroth order Takagi-Sugeno systems. In: IEEE International Conference on Fuzzy Systems (FUZZ), pp. 2013–2018 (2010)Google Scholar
- 16.Kovalev, S.M., Tarassov, V.B., Dolgiy, A.I., Dolgiy, I.D., Koroleva, M.N., Khatlamadzhiyan, A.E.: Towards intelligent measurement in railcar on-line monitoring: from measurement ontologies to hybrid information granulation system. In: International Conference on Intelligent Information Technologies for Industry, pp. 169–181. Springer, Cham (2017)Google Scholar