Support vectors pre-extracting method based on adaptive vector projection

A support vectors pre-extracting method based on adaptive vector projection is proposed. For linear separable problems, the adaptive projection model is constructed, then compute projection line. After the training samples are projected to the line, extract boundary vector sets in one-dimensional space, which are used to train support vector machine(SVM). For non-linear separable problems, the training samples are mapped to high-dimensional space, convert linear separable problems. The orientation of the mean vector is used as the projection line in the feature space. Experiments on two artificial data sets and UCI standard databases show that the proposed method can be as accurate as standard SVM, but is much faster than it.