A multi-layer fuzzy model based on fuzzy-rule clustering for prediction tasks

Fuzzy systems are widely used for solving complex and non-linear problems that cannot be addressed using precise mathematical models. Their performance, however, is critically affected by how they are constructed as well as their fuzzy rule base. Inspired by neural networks that apply a multi-layer structure to improve their performance, we propose a multi-layer fuzzy model with modified fuzzy rules to improve the approximation ability of fuzzy systems without losing efficiency. In practical applications, the fuzzy rule base extracted from numerical data is often incomplete, which makes a fuzzy system less robust. To address this problem, a non-linear function is used as the consequent of each fuzzy rule based on fuzzy-rule clustering to enhance the approximation ability of the fuzzy rule base. In addition, exact matching of fuzzy rules is employed based on the fuzzy rule's antecedent for prediction. By doing so, only one rule will be triggered in each layer, which is very efficient. Experimental results from two simulated functions and three practical applications confirm that our proposed multi-layer fuzzy model can outperform other well-established fuzzy models in terms of accuracy and robustness without sacrificing efficiency.