A new improved support vector machine: QGA-SVM

As a realizing algorithm of the statistical learning theory, Support Vector Machine (SVM) has been paid more and more attention recently. Many researchers have developed some variations of the standard SVM, which is put forward by Vapnik first. SVM has excellent generalization performance compared to the classical learning algorithms, such as NN, ML, etc. In this paper, we developed a new improved SVM algorithm, QGA-SVM. The Quasi Genetic Algorithm (QGA) strategy is integrated in the SVM learning procedure to further optimize and accelerate the training procedure. The tests on some datasets proved its advantages, especially for the multi-class pattern classification with unbalanced classes.