Accuracy improvements in linguistic fuzzy modeling (Q1394745)

The main purpose of the present paper is to provide a comprehensive yet detailed overview of the approaches based on hybrid Soft Computing (SC) systems developed as flexible computing tools to improve the accuracy in linguistic fuzzy systems. The intrinsic need for trading off interpretability and accuracy in the use of SC systems, in particular of fuzzy (logic) systems, is focused in the papers enclosed in the book on improving accuracy while preserving interpretability within fuzzy linguistic modelling. To obtain high degrees of interpretability (i.e. capability to express the behaviour of the real systems in an understandable way) and accuracy (viz, capability to faithfully represent the real systems) is a contradictory purpose and, in practice, one of two properties prevails over the other. While linguistic fuzzy modelling (mainly developed by linguistic fuzzy systems) is centered on interpretability, precise fuzzy modelling (mainly developed by Takagi-Sugeno-Kang fuzzy systems) is concentrated on accuracy. From this perspective, the role played by this book is to present in detail a state-of-the-art on the recent proposals that attempt to obtain linguistic fuzzy models with a good, reliable interpretability-accuracy trade-off inclined however towards improving their accuracy feature. The present book is organized in four main sections comprising (a total of) 15 chapters. Section 1 (Overview) and its only enclosed Chapter 1 (J. Casillas, O. Ordón, F. Herrera and L. Magdalena: ``Accuracy improvements to find the balance interpretability-accuracy in linguistic fuzzy modelling: an overview'') offers an enlightening survey of the various accuracy improvement mechanisms existing in the recent literature. Section 2 (Accuracy Improvements constrained by Interpretability Criteria) collects a set of contributions focusing on the use of different accuracy developments performed under some constraints that avoid an excessive interpretability loss; restrictions such as rigid structures, comprehensibility criteria of the membership functions, or compactness of the rule set are taken into account. This section includes chapters 2--7, whose authors and titles are the following: Chapter 2 (J. Casillas, O. Ordón, F. Herrera: ``COR methodology: a simple way to obtain linguistic fuzzy models with good interpretability and accuracy''); Chapter 3 (F. Cheong, R. Lai: ``Constrained optimization of genetic fuccy systems''); Chapter 4 (H. Ishibuchi, T. Yamamoto: ``Trade-off between the number of fuzzy rules and their classification performance''); Chapter 5 (Y. Jin: ``Generating distinguishable, complete, consistent and compact fuzzy systems using evolutionary algorithms''); Chapter 6 (C. A. Peńa-Reyes, M. Sipper: ``Fuzzy CoCo: balancing accuracy and interpretability of fuzzy models by means of coevolution''); Chapter 7 (J. Valente de Oliveiro, P. Fazendeiro: ``On the achievement of both accurate and interpretable fuzzy systems using data-driven design processes''). Section 3 (Extending the Modelling Process to Improve the Accuracy) contains three contributions that propose more sophisticated modelling processes to attain a good accuracy while preserving interpretability. The involved tiles are: Chapter 8 (C.-Y. Chen, B. D. Liu: ``Linguistic hedges and fuzzy rule based systems''); Chapter 9 (H. Pomares, J. Rojas, J. Gonzáles: ``Automatic construction of fuzzy rule-based fuzzy systems: A trade-off between complexity and accuracy maintaining interpretability''); Chapter 10 (T. Slawinski, P. Krause, H. Kiendl: ``Using individually tested rules for the data-based generation of interpretable rule bases with high accuracy''). The final Section 4 (Extending the Model Structure to Improve the Accuracy) introduces a novel approach that performs the accuracy improvement extending the traditional model structure by utilizing different methodologies such as machine learning, importance factors for each rule, knowledge bases with different granularities etc. The titles of the contained papers are the following: Chapter 11 (E. Aguirre, A. González, R. Pérez: ``A description of several characteristics for improving the accuracy and interpretability of the fuzzy rule learning algorithms''); Chapter 12 (R. Alcalá, O. Cordón, F. Herrera, I. Zwir: ``An iterative learning methodology to design hierarchical systems of linguistic rules for linguistic modelling''); Chapter 13 (P. Carmona, J. L. Castro, J. J. Castro-Schez, M. Laguia: ``Learning default fuzzy rules with general and punctual exceptions''); Chapter 14 (T.-P. Hong, C.-H. Wang, S.-S. Tseng: ``Integration of fuzzy knowledge''); Chapter 15 (L. Sánchez, J. Otero: ``Tuning fuzzy partitions or assigning weights of fuzzy rules: which is better?''). The special merit of this volume is that of being really useful at least from two points of view: (a) for non-expert readers, whatever their background would be, to easily get a good knowledge about this area of expertise, and (b) to provide a true support for specialists wishing to discover the latest results and trends in fuzzy modelling research.