Research and application on convex-hull-SVM based on Two-Phase method

Support vector machine (SVM) becomes a research focus in the field of machine learning for it is based on expected risk minimization and overcomes “the curse of dimensionality” effectively. In the theoretical research, the focus is mainly on the choice of kernel function and the fast algorithm. On this background, a convex hull algorithm based on Two-Phase method is proposed as one of fast algorithm in this paper, which solves the equation of convex-hull fast by two-phase method so as to reduce the input sample points and then improves the efficiency of the implementation of SVM. Iris simulation shows that the new algorithm is effective and feasible. What´s more, the improved convex-hull-SVM is applied to server performance alarm, and a performance alarm core algorithm based on which is proposed to classify the server status quickly and accurately..