Identification and Convergence Analysis of Interval Type-2 Takagi–Sugeno–Kang Fuzzy Systems for High-Dimensional Classification Problems
Qinwei Fan, Deqing Ji · IEEE Transactions on Fuzzy Systems · 2025
In this paper, a new de-fuzzification algorithm is proposed for multi-classification problems, which can effectively improve the accuracy, stability and computational efficiency of interval type-2 fuzzy systems. In addition, in order to enable the fuzzy system to handle high-dimensional data, this paper also designs a collaborative feature selection strategy based on gate function and GroupL0regularisation, which effectively solves the challenges faced by fuzzy systems when dealing with high-dimensional problems. The strategy allows the system to select relevant features and alleviate the curse of dimensionality. Finally, we employ a Root Mean Square Propagation algorithm to simultaneously optimise the antecedent and consequent parameters in the interval type-2 TSK (Takagi-Sugeno-Kang) fuzzy system, and conduct a convergence analysis of the algorithm to ensure the validity and reliability of the proposed method. To verify the performance of the proposed algorithm, we conducted simulation experiments on high-dimensional datasets. The results demonstrate the superiority of our method in handling multiclassification tasks and the ability to handle complex highdimensional data.