What I'm wondering with a markov process is whether it could be embellished to include other potentially relevant variables. From 5 minutes reading wikipedia, it seems like I'd have a combinatorial explosion of states, and the more states, the more data needed to train the model.

So I'd have like 48 states, for each half-hour of the day, times 3-4 for 8-11 hours long sleep? Would it work to have ordered pairs where the first item is measured in time since my last awakening?