Multiple Bayesian Sensors

Configuration
1

I’m trying to figure out how to add multiple Bayesian sensors. I’d like to track when I am home and when my wife is at home. This is the yaml code that I thought would work, but I get a duplicate key error:

binary_sensor:
  # Bayesian sensor determining if I am home
  - platform: bayesian
    name: "Me_home"
    prior: 0.75
    probability_threshold: 0.75
    observations:
      - entity_id: "device_tracker.my_phone"
        prob_given_true: 0.75
        prob_given_false: 0.25
        platform: "state"
        to_state: "home"

binary_sensor:
  # Bayesian sensor determining if Wife is home
  - platform: bayesian
    name: "Wife home"
    prior: 0.60
    probability_threshold: 0.75
    observations:
      - entity_id: "device_tracker.wifes_phone"
        prob_given_true: 0.75
        prob_given_false: 0.25
        platform: "state"
        to_state: "home"

Screenshot 2025-05-21 191613
Screenshot 2025-05-21 191613565×615 30.4 KB

2 
binary_sensor:
  # Bayesian sensor determining if I am home
  - platform: bayesian
    name: "Me_home"
    prior: 0.75
    probability_threshold: 0.75
    observations:
      - entity_id: "device_tracker.my_phone"
        prob_given_true: 0.75
        prob_given_false: 0.25
        platform: "state"
        to_state: "home"

  # Bayesian sensor determining if Wife is home
  - platform: bayesian
    name: "Wife home"
    prior: 0.60
    probability_threshold: 0.75
    observations:
      - entity_id: "device_tracker.wifes_phone"
        prob_given_true: 0.75
        prob_given_false: 0.25
        platform: "state"
        to_state: "home"
3

I tried that. “Wife home” does not show up as an entity. But “Me_home” does show up as an entity.

4

What I posted is correct.

5

It is indeed. I changed the names and it worked. Thank you.

6

I always have trouble getting my head around the Bayesian integration, and you may just be experimenting, but how did you work out your prob_given_false etc?

If I’ve understood correctly, when your phone’s home there’s a 75% chance that you are too; when your phone’s not home, there’s a 25% chance that you’re home. That suggests that you leave it behind a lot, and that it’s “unavailable” a lot of the time (I think…).

Mine is:

    # Phone home (Bluetooth LE)

    - platform: "state"
      entity_id: device_tracker.nokia_bermuda_tracker
      to_state: "home"
      prob_given_true: 0.95
      prob_given_false: 0.01

…and I do leave it behind occasionally.

One of the great things about Bayesian in HA is that there’s lots of historical data to use. There’s an interactive spreadsheet here to save you racking your brain over the maths:

Also, the more observations you have, the better it works!