Identifying an automaton model for timed data

A model for discrete event systems (DES) can be learned from observations. We propose a simple type of timed automaton to model DES where the timing of the events is important. Learning such an automaton is proven to be NP-complete by a reduction from the problem of learning deterministic finite state automata (DFA) without time. Based on this reduction, we show how the currently best learning algorithm for DFAs (state merging) can be adapted to deal with time information.