Parameter Selection for Twin Support Vector Machines Based on Improved Tuna Swarm Optimization Algorithm

The Twin Support Vector Machine (TWSVM) is a classification algorithm derived from the Support Vector Machine (SVM), exhibiting notable efficacy in binary classification problems. However, TWSVM is not without its limitations. In TWSVM, the selection of different penalty coefficients significantly influences the classification performance across diverse datasets. Inappropriate parameters may lead to a substantial degradation in the classifier's performance. Addressing the challenge of penalty coefficient and kernel function parameter selection in TWSVM, this study introduces the STSA-TWSVM algorithm. Firstly, a Sobol sequence is employed to uniformly generate initial positions for the population in the target solution space. Subsequently, a global optimization is conducted using the Tuna Swarm Optimization algorithm. Finally, comparative experiments are carried out on the UCI dataset against the TWSVM based on the other Optimization algorithms. The experimental results indicate that the proposed STSA-TWSVM algorithm enables more accurate classification.