Problem: After devices are “learned” Sense owners need to determine what real-world devices they map to. One way to do that for devices is using a time series approach/view.
For example, if we know that real-world appliance A (say, a dryer) was on from 3pm to 6pm today, being able to view the set of learned devices (say, three learned devices that Sense thinks are all dryers) during the time we know the dryer was on helps validate or invalidate Sense’s determination. If any or all of them are active inside the time window, but not outside, that counts as evidence validating the match. If any or all of them are inactive inside the time window, that counts as evidence invalidating the match (obviously confidence goes up as multiple observations are recorded). The negative case is also useful, when a learned device shows active load when we know that the real-world device is powered off.
In our example, it would be useful to be able to add and remove multiple learned devices to/from a time series view to see how they line up with our ground truth knowledge of what devices were actually running or not running at that time. A follow-on requirement to this would be the ability to make some kind of notes on learned devices to note what has been learned.
For example, if I’ve been able to prove that a specific learned device is not the furnace, the fridge, the dehumidifier, etc. using time series or any other method, there should be a way to record and remember that on the learned device in the Sense apps.