Title:Granular Ball K-Class Twin Support Vector Classifier

Abstract:This paper introduces the Granular Ball K-Class Twin Support Vector Classifier (GB-TWKSVC), a novel multi-class classification framework that combines Twin Support Vector Machines (TWSVM) with granular ball computing. The proposed method addresses key challenges in multi-class classification by utilizing granular ball representation for improved noise robustness and TWSVM's non-parallel hyperplane architecture solves two smaller quadratic programming problems, enhancing efficiency. Our approach introduces a novel formulation that effectively handles multi-class scenarios, advancing traditional binary classification methods. Experimental evaluation on diverse benchmark datasets shows that GB-TWKSVC significantly outperforms current state-of-the-art classifiers in both accuracy and computational performance. The method's effectiveness is validated through comprehensive statistical tests and complexity analysis. Our work advances classification algorithms by providing a mathematically sound framework that addresses the scalability and robustness needs of modern machine learning applications. The results demonstrate GB-TWKSVC's broad applicability across domains including pattern recognition, fault diagnosis, and large-scale data analytics, establishing it as a valuable addition to the classification algorithm landscape.