Dual center based intuitionistic fuzzy plane based classifiers

Anuradha Kumari, Muhammad Tanveer

Abstract: The plane-based classifiers, support vector machine (SVM) and twin support vector machine (TWSVM), are susceptible to the negative impact of noise, outliers, and class imbalance learning (CIL). The concept of intuitionistic fuzzy (IF) addresses the impact of noise and outliers present in the data. However, IF is sensitive to the threshold distance value and calculates the distance of each data point to other points of its class which is computatinally intensive. Moreover, IF assigns the same score value to the data points that have distinct distances from the center of the other class and identical membership values. To address the aforementioned limitations, in this paper, we propose dual center-based intuitionistic fuzzy plane-based classifiers, specifically dual center-based intuitionistic fuzzy support vector machine (DC-IFSVM) and dual center-based intuitionistic fuzzy least square twin support vector machine (DC-IFLSTSVM). The proposed models, DC-IFSVM and DC-IFLSTSVM handle the CIL by associating class-specific terms in the weighting function. The optimization problem of DC-IFLSTSVM is solved using the conjugate gradient (CG) method. We conducted extensive numerical experiments of the proposed DC-IFSVM, DC-IFLSTSVM and baseline models on 54 UCI and KEEL datasets, which resulted in the superiority of the proposed DC-IFLSTSVM. Furthermore, experiments on the BreakHis dataset, focused on classifying benign tumors from malignant tumors, showcased the remarkable performance of the proposed DC-IFLSTSVM. The code for the proposed model can be found on https://github.com/mtanveer1/DC-IFLSTSVM.