All program have some probabilities of failing. It may be small but nevertheless it is present. Embedded system is usually a critical system and there can be many parts which is error prone. Usually errors are predicted by change history (problem fix, requirement change etc). That is, which part is changed frequently, contain more error, that's the principle.

Is there any other way to predict error? If i do not have change history records then what can be other metrics? Can tester/software manager use their experience to predict which parts are error prone? For example, embedded system handles different types of resources. So those places are needed to be checked where resources are handled. Because from our experience those parts contains more error.