The construction of a support vector machine using the shuffled frog-leaping algorithm

摘要

The parameter setting of a support vector machine is a very important factor to its accuracy and efficiency. In this paper, we employ the shuffled frog-leaping algorithm (SFLA) to simultaneously train all parameters of the support vector machine including the penalty parameter, smoothness parameter and Lagrangian multiplier. The proposed method is called the SFLA-based support vector machine (SFLA-SVM). In experiments, binary and multi-class classifications are explored. In experiments, 10 of the benchmark data sets of UCI Machine Learning Repository are used. The classification performance of SFLA-SVM is compared to the original LIBSVM method associated with the grid search method and the PSO-SVM. The experimental results advocate that the use of SFLA-SVM for classifying the pattern classifications has better classification accuracy.