GBTWSVM: Granular-Ball Twin Support Vector Machine

Twin Support Vector Machine (TWSVM) has gained popularity as a machine learning tool due to its low computational complexity. However, it may not be the most adaptable to all types of datasets containing noise and outliers. Granular-ball computing (GBC), known for its higher robustness to noise and outliers, provides a potential solution. In this paper, we introduce the Granular-ball Twin Support Vector Machine (GBTWSVM), which combines the strengths of GBC and TWSVM. We optimize TWSVM by using coarse-grained granular-balls as inputs instead of individual samples. Subsequently, we develop the dual model of GBTWSVM and design an algorithm for classifications. Through numerical experiments conducted on seventeen benchmark datasets, we demonstrate that GBTWSVM outperforms TWSVM, Intuitionistic Fuzzy Twin Support Vector Machines (IFTSVM) and Granular-ball Support Vector Machine (GBSVM) in terms of running time, accuracy, precision, and recall.