I’m looking to implement some bayesian sensors to determine if certain rooms are currently occupied. As I’m designing it, I keep thinking of more observations to add (motion sensor recency, time of day, home/away for multiple people, whether my computer is on, etc). It just keeps getting bigger and I want to know this will actually work.
Does anyone have a large bayesian network running with multiple sensors stringed together? Does it work well? Any tips on how to design a network or come up with the probability estimates?
I know bayesian network normally assume an acyclic graph, but I’ve done testing and HA allows cycles. Is anyone running a cyclic graph? Are there any potential pitfalls to setting one up?