Thermonuclear fusion devices of the Tokamak type can operate in distinct confinement regimes, which present different properties in terms of performance and plasma parameters. Discriminating among them in real time would represent a useful advantage for an efficient control of the experiments. A comparison between two automatic identifiers, one based on fuzzy logic and another based on classification and regression trees, is presented. A robustness assessment, adding Gaussian white noise to the input signals, was performed to determine the properties of the two approaches in realistic experimental conditions.