I have a project in mind to use machine learning to detect that a home automation sensor has failed.
It's very possible to just write a set of rules in JS to determine whether the sensor is working or not, but I'd like to delve a little into ML.
Has anyone tried this before?
Can you point me in the right direction?
I have been playing with the machine learning v2 node, but it looks unmaintained. I have installed all the prerequisites and will just get on with reading about each component until I figure something out.
EDIT: I get you need quite a bit of data. I don't think this is an issue as sensor data over a period of time adds up and when you have a hundred of them it doesn't take long at all to accumulate a data set.
Although I'm aware that not all types of sensors can be predicted, it would be very useful to have something like you explain in Node-RED.
At my daily job we host a large series of servers.. Since last month we are using Dynatrace for monitoring both our servers and our applications. Dynatrace uses AI for calculating a baseline (based on what Dynatrace has learned in a moving window of the last 7 days), and uses that knowledge to detect abnormalities in the metric values of today. I must admit that I have become a huge fan of this kind of stuff. It is absolutely not always correct but it is a huge help: it is like you suddenly get 20 pairs of extra eyes to watch your graphs. Very useful.
Hopefully somebody with knowledge in this area can assist you here..
Bart
Could this article be of any help to you? His Github repository seems to be maintained recently. Moreover it is written in javascript, and based on tensorflow.js. So you should be able to run it with Node-RED, without having to install third party stuff...
Perhaps you can contact the author, if you don't know how to get started with it...
