We applied signal processing techniques and mathematical methods to the examination of the prognostic power of 210 descriptive parameters extracted from resting-state EEG patient data obtained at admission to intensive inpatient neurorehabilitation. Patients with unresponsive wakefulness syndrome (UWS) and patients in a minimally conscious state (MCS) were classified into binary groups by fitting a generalised linear model on training data and examining the model on test data. We aimed subsequently to develop an automated system for outcome prediction in severe disorders of consciousness (DOC) by selecting an optimal subset of features using sequential floating forward selection (SFFS). Several parameters performed significantly better than randomly permuted feature vectors at predicting outcome and/or discriminating DOC states. The combination of the optimal subset of features selected with SFFS seem to afford high prediction power.