Linguistic fuzzy rule learning through clustering for regression problems

Abstract

The fuzzy rule-based system model is an important and active research line in the fuzzy logic community looking for compact and robust systems with a high level of accuracy-interpretability of trade-offs. On the other hand, fuzzy rules-based systems provide accurate and interpretable solutions that give the ability to handle Complex data and uncertainty. It has also been historically applied in the solution of classification and regression problems. In this paper, the authors present a new fuzzy approach for solving problems of regression based on linguistic fuzzy rule learning with subtractive clustering and linguistic modifiers. The proposed system includes two phases for getting linguistic fuzzy rules: Multi-granularity, fuzzy discretization of the linguistic variables and linguistic approximation of fuzzy rules learned. Regarding experiments, researchers used twelve real-world data sets to compare the proposed system with three of the most widely used simplified fuzzy genetic systems: FSeMOGFS+TUNe, A-METSK-HDe and FRULER. The results highlight the competitiveness of the model in terms of accuracy and its superiority in interpretability.