Modelling of Data-Driven Monotone Zero-Order TSK Fuzzy Inference System
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
UNIVERSITI MALAYSIA SARAWAK
Abstract
Description
In this thesis, the use of two data-driven Fuzzy Inference System (FIS) model, i.e., ad hoc and system identification methods in the literature are analyzed. The data-driven FIS model is used to develop monotone fuzzy If-Then rules based on single and multi-attribute data for modelling a zero-order Takagi-Sugeno-Kang (TSK) FIS model. Convex and normal trapezoidal (triangular) fuzzy sets coupled with a strong fuzzy partition strategy are employed. Firstly, a monotone TSK FIS model that can always be obtained using a set of complete and monotone fuzzy If-Then rules is explained. A monotone test is devised to evaluate the degree of monotonicity (DOM) of TSK FIS models. The analysis further indicates that with single-attribute monotone data, monotone fuzzy If-Then rules can be produced using both ad hoc and system identification methods. However, this is not the case for multi-attribute monotone data. This finding is important for developing a monotone TSK FIS model, as it highlights that even with a “clean” monotone data set, the generated fuzzy If-Then rules may need to be pre-processed (i.e., re-labeled), in order to guarantee the monotonicity property of the resulting TSK FIS model. As such, a framework for designing and developing monotonicity-preserving FIS models for ad hoc and system identification methods are proposed, respectively. The effectiveness of the proposed approach is evaluated using a number of simulated examples and a real-world benchmark problem. The results indicate that the proposed approach coupled with a monotone data set leads to a lower mean square error and a better DOM score.