Since the regression trees provide means to identify the variables which influence most strongly the severity of a contingency, they may suggest which preventive control alternatives the operator could apply to reduce the latter severity. However, at the present stage it is not clear how this could be achieved so as to actually increase security, i.e. without leading to simultaneous reduction of pre-contingency LPM for example.
On the other hand, once the most dangerous events or sequences of events have been identified, more refined simulations may be applied to pre-determine possible corrective control actions, e.g. according to the approach proposed in [7].