Analyses of ITER operation mode using the support vector machine technique for plasma discharge classification

A new approach is proposed for classifying tokamak plasma discharges. The method is based on a modern data mining technique—the so-called 'support vector machine', which is able to construct the optimal classifier. The international database of plasma discharges from different tokamaks has been analyzed with respect to H- and L-modes. A new linear equation, which separates H- and L-modes in the space of eight parameters, was obtained allowing us to classify a tokamak pulse as H- or L-mode and to give a quantitative estimate of how deep the pulse is in a mode. The equation also allows calculation of the value of the H-mode threshold for a selected plasma characteristic. The results are applied to ITER parameters. It is shown that in the main regimes ITER should operate deeply in H-mode. A more optimistic than known H-mode loss power threshold prediction is obtained for ITER.