Do Sense device models dynamically update?


#1

Once Sense detects a device and a model is loaded into the monitor, is it “tweaked” as time progresses? I have a few devices that have been ID’d, but aren’t just right. For example, my heat pump water heater only has about half of its usage detected, and my air conditioner has been replaced, but the old Sense model is still being applied to it, which doesn’t account for 2 compressor stages in the new system. Will Sense figure these things out on its own as time goes on?


#2

Was wondering the same…

I look forward to this discussion. The answer to your question will determine whether the best option is to “wait it out” and allow Sense to tweak versus Deleting the Device and starting over.


#3

So, based on my experience, deleting the device just seems to remove the model from your local monitor. The back-end learning algorithm that seems to operate in the sense “cloud platform” continues to analyze your historical data and will regenerate the model and re-add the deleted device to your app (and push the model back to your monitor) even if it is not present anymore. I have a bunch of tools that a contractor was using in my house that, even when I delete them, show back up again a few days or weeks later as newly found devices. I’ve opened a ticket with Sense and they claim to be working on the problem but say it’s not a quick and easy fix.


#4

Models definitely do get updated as more data is accrued. This is the reason that data resets don’t necessarily net the same exact detections from before the reset…models have been altered and are firing from revised triggers. And yes, deleting the device just removes it locally, but it also alerts us that it was deleted and thus gives the data science team another data point on which to tweak the detector. The timeline of this process varies, of course, so some models will be revised faster than others.