Binary sensor that detects data patterns (new component)

I recorded my washing machines power usage throughout the “Cottons” washing cycle which gave me an interesting set of data points shown here:

Has there been any work done in the data pattern detection space? I can see there are statistics and history_stat sensors, which is fair enough, but nothing that does pattern detection.
Crazy to think that we have machine vision solutions like movement detection for cameras, yet no pattern detection for data.

The sensor would take a data set as input and compare it to another sensor that records power usage for example. As the difference between data set and power sensor is aligned to the changes seen in the dataset, we can turn on a binary_sensor to show that the data set matches.

This is not instantaneous, the user could supply a confidence value. If 20% of dataset (thus far) matches, trigger binary sensor to on.

This has lots of applications I imagine…beyond detecting which bloody cycle your washer is running! :slight_smile:
What are your thoughts?

5 Likes

Hi @danny,

Did you ever manage to find something which would do as you wanted? I am just trying to do something similar, but would like to even get down to detecting sections of the wash cycle (i.e. idle, fill, wash, spin, dry, finished).

No I gave up on this. It was just an interesting problem at the time.