Timed Automata for Behavioral Pattern Recognition

We argue that timed models are a suitable framework for the detection of behavior in real-world event systems. A timed model which detects behavior is constructible by a domain expert. The inference of these timed models from data is a hard problem. We prove the inference of a class of timed automata (event recording automata) to be harder than the inference of finite automata.Software Computer TechnologyElectrical Engineering, Mathematics and Computer Scienc