Automatic detection of L-H transition in KSTAR by support vector machine

Method for automatic detection of L-H transition using Support Vector Machine (SVM), a popular tool of supervised machine learning tools, has been evaluated in order to improve plasma density control in KSTAR. Through the SVM, a nonlinear classifier is trained to distinguish L-mode and H-mode using two kinds of diagnostic data measured in KSTAR. The trained classifier has been analyzed for possible usage on the real-time detection through the truncation of the training samples. Study on the optimization of the training samples, and corresponding accuracy change is made for evaluating feasibility for real-time implementations.