A new method to improve the sensitivity of support vector machine based on data optimization

As a new type of learning machine based on statistical learning theory, SVM has been extensively applied in some topics of machine learning. For many applications in the pattern recognition, the classifier is desirous to have a higher sensitivity. Considering that the current methods for improving the sensitivity of SVM possess some deficiencies, we present a new method based on data optimization in this paper. Its scheme is to control the sensitivity of SVM by optimizing the training data with other statistical model. Test results with the identification of translation initiation site of Eukaryotic gene show that data optimization based method can improve the sensitivity and overall prediction accuracy of SVM effectively, and composed with other method will get better effect.