How Does Sense Machine Learning Deal With Data and Concept Drift?

Cross-posting this Sidarth Singh question and Sense answer from the Sense Facebook group.

Sense uses machine learning to understand the signature of appliances and identify them automatically. But if the model performance itself degrades (due to data drift) , how will sense make sure of that situation not happening ?

Hi Sidharth Singh - Sense models run iteratively on the past 60 days of data for the home, allowing each model to update according to the most recent behavior of the device. For example, if you started running a space heater twice as long as you normally do, the expected duration (duration model) would adjust according to the most recent behavior.

2 Likes

I find it impossible not to mention Rumsfeld’s “known knowns”.

Somewhat aside, regarding:

I’m again pondering ML on the Solar CTs.
Panels degrade; snows falls; trees shadow; birds poop.

This leads also to ML on any dedicated device CT.
Water heater elements degrade and HVAC compressors misbehave, yes.
I have dedicated CTs on a boiler circuit that includes 5 distinct “devices”: oil pump; furnace fan; 2 water pumps; and a light to the boiler room.

2 Likes

Sense models run iteratively on the past 60 days of data for the home, allowing each model to update according to the most recent behavior of the device. For example, if you started running a space heater twice as long as you normally do, the expected duration (duration model) would adjust according to the most recent behavior.

1 Like