Old thread but still interest me. What I’ve done is; My example, occupancy and motion. A convoluted presence detection.
Devices, all Aqara entry T1s? and P1 motion sensors, about 7. More evidence “may” be best. And Reolink person detected and a convoluted cloud trick for Wyze cameras for person seen.
How an Aqara motion sensor works from factory. Motion triggers, clears in 30 seconds.
Occupancy also triggers at the same time as motion.
Occupancy takes 10 minutes to clear and if new motion occurs within that 10 minutes, it resets and starts the count down at 10 minutes, continuously with new motion. Can hold my Bayesians thresholds for 10 minutes or more.
Retirement fun, lol, 
I created several mini Bayesians.
Computer room, bedroom, basement, trash day, garden, outside. Etc.
Computer room is used heavily in the morning. Path of travel concepts.
Hallway motion sensor see someone coming, ok, going to bedroom or computer room? So hallway some evidence but not a big spike.
Computer room motion detects, occupancy fires on, while there, unless absolutely still, motion will continuously trigger occupancy.
For my sensors, it’s safe to assume someone is in the computer room, so motion and occupancy are both in the observations. Holds Bayesian above the threshold.
Leaves computer room? Motion clears in 30 seconds, occupancy clears in 10 minutes.
(These Aqara motion device settings for at least motion are adjustable) occupancy? I’d have to look.
I have gauge’s for each Bayesian to show probability, template a sensor helpers to display probability on the gauges.
History graphs show room occupancy, and motion sensors display on another entities card.
Trash day;
Pulls Trash keyword in Google calendar events, automation to remind me, take trash down, two reminders, 11AM & 3PM.
Only on Tuesday. Ignored holidays and snow days for now. Wednesday also looked at in events, but manual entry of keyword (for now).
If completed, another automation sends notification that trash bin moved at so and so time. Etc.
So the Trash Bayesion is looking at on Tuesdays, during 11AM_5PM, has inner garage door opened, has Wyze V3 seen a person, has right outer garage door opened.
Has driveway Wyze V4 camera seen a person, walks down trash can, checks mail box, did my phone exit the home zone? Should have. Being seen by cameras coming back, right garage door closes, entry goes to off. Big spike trash has been completed. All observations in the Bayesian.
This would become a tedious post if I continued much further.
In case someone reads it and doesn’t know, visit the HA Bayesian page, it has a GUI integration for easier GUI building of the Bayesian sensors.
Read the document above, build a small Bayesian, think thru travel paths of family, time, for us old folks, it’s just two people.
While looking to load the integration, look at the sample Bayesians on the webpage and read it thoroughly.
It’ll take some time, effort and frustration, maybe some Ai assistance. Use Ai with a grain salt, trust but verify.
Good luck, I’m still tweaking Bayesians.
But, understanding Bayesians more & more every day. Have to build some more.