The Research And Optimization On Support Vector Machines Algorithm

The support vector machines (SVM) is a new universal machine learning method based on statistic theory and proposed by Vapnik. In recent years, SVM has been given considerable attention due to its excellent generalization performance in a wide variety of learning problems, such as handwritten digit recognition, classification of web pages and face detection. However, being a new technique, there still exist some open questions that should be deliberated by further research.When the amount of training samples is too large, this method has many shortcomings such as the computation time,the memory expenditure and the computation accuracy. In order to solve this problem and improve the speed of training SVM, a kind of algorithm for SVM is proposed. In order to solve problem of SVM parameters, immune algorithm applied to model selection of SVM. This algorithm can improve the classification accuracy of SVM. SVM has a poor performance, when the two-class problem samples are very unbalanced. To significantly improve the classification performance of imbalanced datasets, a kind of algorithm for the unbalanced samples is proposed. The experiments on the UCI database are done with those algorithms. Experimental results indicate those algorithms are particularly effective.In the last, we summarize the paper's contents and propose some suggestions of the future work.