Multiple Support Vector-based approach to Breast tumor recognition

摘要

Opisano wykorzystanie metody MSVM (multi-support vector machinę) do wykrywania guza piersi. Metoda generuje klasyfikator wektorowy w ośmiu kierunkach a selekcję cech przeprowadza się wykorzystując funkcje Gaussa jako kernel.%Based on Multi-Support Vector Machine (MSVM), this paper provides a method of MSVM for breast tumor recognition to solve unfixed size and individual difference with the breast tumor. Support Vector Machine (SVM) on the eight direction of bump area is taken to generate vector classifier and to select Gauss kernel function as kernel function. The system application and test shows that the MSVM in breast tumor recognition achieved good result, and provide the reliable basis for further medical diagnosis. The breast tumor recognition accuracy achieved 97.3% when o=30.