Interval Type-2 Fuzzy-Support Vector Regression in Representation of Uncertainty in a Non-linear System

Abstract

Machine learning algorithms such as Support Vector Machine (SVM and Support Vector Regression (SVR) are faced with challenges when confronted with imprecise and noisy data, which can lead to less meaningful outcomes. This paper introduces Interval Type-2 Fuzzy Support Vector Regression (IT2FSVR) as a solution to address uncertainty in non-linear systems. By combining Interval Type-2 Fuzzy Sets (IT2FS) and SVR, the proposed method enhances performance in systems with high levels of noise and non-linearity. The integration of IT2F membership in SVR directly tackles uncertainty in prediction problems, enabling adaptive learning to varying inputs and improving generalization performance. To demonstrate the effectiveness of this approach, the authors tested the performance of IT2F-SVR using a dataset of cardiovascular disease patients. Experimental results demonstrate that IT2F-SVR effectively eliminates uncertainty and significantly improves the learning process, outperforming individual approaches when applied to the same dataset and achieving faster execution times compared to some alternatives, albeit taking more time than SVR. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.