Monotonic relation-constrained Takagi-Sugeno-Kang fuzzy system

The Takagi-Sugeno-Kang fuzzy system has wide applications across different areas, e.g., regression, classification and decision making, attributed to its high precision and inter-pretability. However, the existing Takagi-Sugeno-Kang fuzzy system is not an ideal solu-tion to some special scenarios, particularly for those that are constrained monotonically. To this end, a monotonic relation-constrained Takagi-Sugeno-Kang fuzzy system classifier is proposed in this paper. The proposed method introduces a monotonic relation between the inputs and the outputs, where the objective function is expressed in a monotonically constrained form and a strategy for generating monotonicity constraint pairs is developed. Furthermore, to address the convexity loss caused by the increasing monotonicity con-straints, the proposed method introduces the Tikhonov regularization strategy to ensure the uniqueness and boundedness of the solution. The results from extensive experiments show that the proposed method exhibits better classification performance than the original Takagi-Sugeno-Kang fuzzy system and state-of-the-art monotonic classification methods in handling monotonic datasets. (c) 2021 Elsevier Inc. All rights reserved.