Intuitionistic Fuzzy Twin Kernel Classifier Improves Computing Speed by 52 Percent

A novel machine learning model known as the dual centroid-based intuitionistic fuzzy twin kernel ridge regression classifier, abbreviated as DC-IFTKRR, has achieved an average classification accuracy of 93.92 percent while executing approximately 52 percent faster in mean computational speed than an earlier dual centroid architecture. Detailed in a regular research paper published on 13 November 2026 in the International Journal of Data Science and Analytics, the investigation was authored by Barenya Bikash Hazarika, Deepak Gupta, Parashjyoti Borah, and Ma. Louella M. Salenga. The study focuses on resolving key computational trade-offs in pattern classification.

The foundation of the study is grounded in the well-known universal generalization capability of the powerful kernel ridge regression (KRR) classifier. In recent machine learning developments, intuitionistic fuzzy (IF) sets have been combined with kernel KRR, twin support vector machines (TSVM), extreme learning machines, and random vector functional link models to deliver excellent classification performance across multiple settings. However, the researchers identified that substantial scope remained for further enhancing generalization performance while concurrently preserving a decent computing speed during model execution.

To pursue enhanced generalization, the investigators initially examined a dual centroid (DC) based intuitionistic fuzzy twin bounded large margin distribution machine, termed DC-IFTBLDM. Despite its conceptual merits, the researchers reported that the major flaw of the DC-IFTBLDM model was that it proved computationally very expensive to run. The heavy computational burden stemmed from the structural necessity to solve large quadratic programming problems (QPPs), which created substantial computational overhead and limited execution efficiency when processing data. Consequently, there remained a clear necessity to devise an alternative mathematical formulation that could avoid solving large quadratic programming problems.

To overcome the computational expense associated with solving large quadratic programming problems, the authors proposed the novel DC-based intuitionistic fuzzy twin kernel ridge regression (DC-IFTKRR) classifier for efficient classification. Rather than solving a quadratic programming problem, DC-IFTKRR is structured to solve a set of linear equations. This fundamental mathematical shift directly reveals the model’s time-efficiency, enabling the classifier to attain its approximately 52 percent faster mean computational speed when compared directly to the DC-IFTBLDM framework while maintaining strong predictive performance.

To confirm the validity of the proposed DC-IFTKRR model, the research team carried out extensive comparative experiments against seven existing algorithms across several interesting benchmark datasets. The baseline algorithms selected for the comparative evaluation included standard twin support vector machines, classical kernel ridge regression, intuitionistic fuzzy kernel ridge regression, intuitionistic fuzzy twin support vector machines, weighted kernel ridge regression, intuitionistic fuzzy broad learning systems, and the earlier DC-IFTBLDM framework. Across these evaluations, the high generalization performance of DC-IFTKRR was evident from the empirical results.

In addition to evaluating classification accuracy across the various benchmark datasets, the authors documented the parameter sensitivity of the DC-IFTKRR classifier.

Finally, the researchers presented a real-world application on a leaf dataset to reveal the practical efficiency of the DC-IFTKRR model when applied to real-world problems. In this real-world application, the proposed classifier demonstrated the highest classification accuracy among the tested methods, attaining an accuracy rate of 90.03 percent on the leaf dataset. The experimental findings indicate that combining dual centroid intuitionistic fuzzy principles with twin kernel ridge regression offers a viable approach for achieving accurate and computationally efficient classification.

Subject of Research: Dual centroid intuitionistic fuzzy twin kernel ridge regression classification

Article Title: Dual centroid-based intuitionistic fuzzy twin kernel ridge regression classifier

Article References: Hazarika, B. B., Gupta, D., Borah, P., & Salenga, M. L. M. (2026). Dual centroid-based intuitionistic fuzzy twin kernel ridge regression classifier. International Journal of Data Science and Analytics, 22(1), Article 385. https://doi.org/10.1007/s41060-026-01344-x

Image Credits: AI Generated

DOI: 10.1007/s41060-026-01344-x

Keywords: kernel ridge regression, intuitionistic fuzzy sets, twin support vector machines, extreme learning machine, random vector functional link, dual centroid, large margin distribution machine, DC-IFTKRR, DC-IFTBLDM, quadratic programming problems, linear equations, computational speed, classification accuracy, leaf dataset, parameter sensitivity, benchmark datasets, generalization performance, weighted KRR, broad learning systems