An improved QDPSO training hypersphere one class support vector machine

Combined with the optimization strategy of hypersphere OC-SVM and QDPSO, an improved QDPSO method is presented to training hypersphere OC-SVM. In this method, the new position of the directional particle is calculated based on the current global best point(gBest), which optimized direction conforms to Zoutendijk fastest decline method principle. In the initialization, the position of one particle is initialized according to SMO, which make its position nearer to the global optimum solution; and the boundary points of subjected plane are concerned as the initialized position of other particles their, so as to make the searching area wider. The experiment shows, the convergence and the generalization of D-QDPSO is good, the misrecognition of D-QDPSO is 0.12% lower than that of SMO, the operating speed is 2 times faster than that of LPSO.