Units Device Service Machine Learning Module

This is a set of ML solutions and functions used in Mostat project.

Currently there are implemented couple of ML routines, see below:

• Predict 2x AA battery chemical composition used in Mostat room sensor and other AG products

Predict battery chemical composition

The idea is to predict the chemical composition of the battery according to its voltage and age. As you know we using Room Sensor device with 2x AA type battery. In our case the battery chemical composition may be two types Lithium (Energizer Ultimate, etc..) and Alkaline (Duracell Turbo, etc..). The battery age starts from battery first registration date and the age is counts in days, so first registration date should be 0 days. The above information is used to precisely calculate battery percentage. In other world, we make battery voltage range representation in percents:

• Lithium range from 3.400V and above - 100%, 2.600V - 0%
• Alkaline range from 3.100V and above - 100%, 2.300V - 0%

For prediction we are using neural network library synaptic.js. The architecture is multilayer perceptrons. See Perceptron for more information.

There is 2 input layers - voltage and age. Two hidden layers and one output layer for chemical composition.

The UDS ML is used in Mostat Wi-Fi thermostat produced by Alien Green LLC

Installation

Installing from github like this:

npm install aliengreen/uds-ml

Usage

To predict battery chemical composition you should call function predictBatteryChemicalComposition and pass two parameters. Current battery voltage (for 2x AA battery) e.g 2.987 and age in days e.g. 60.

Below is an example how to use this function.

var udsml = require('uds-ml');

let result = udsml.predictBatteryChemicalComposition(2.987, 60);

console.log(result); // 'Alkaline'

Possible answers:

• Lithium
• Alkaline
• Unknown

Notes

@TODO

License

Licensed under The MIT License (MIT)
For the full copyright and license information, please view the LICENSE.txt file.