[R] Recognizing patterns in time series data

James Courtier-Dutton james.dutton at gmail.com
Fri Feb 28 11:00:04 CET 2014

See the attached picture.

It shows the temperature of a water pipe in a heating system.
Notice the up/down/up/down bit. This is the boiler turning on and off
again to keep the water at a particular temperature.
Are there any R libs/modules that would help me recognizing patterns
like this in graphs?
I have looked through the various R libs/modules and found lots of
statistical analysis tools, but I want pattern recognition and not
statistical tests.

I.e. So I can tell when the boiler was fired up, and when it was waiting.
I also want to use to graph to estimate how much time the boiler was
actually lit and heating water, to get an idea regarding gas usage,
and then compare that to the gas meter for calibration purposes.
I would also like to recognize other features of the graph, like any
unexpected changes in temperature, due to someone running a hot tap

For those interested, I have made my own heating/hot water controller
and monitoring setup with an arduino, DS18B20 digital temp sensors,
and relays for switching values on/off and sending "call for heat" to
the boiler.
The graph is the output from one of the DS18B20 one wire temp sensors.
For safety, I am not connecting anything to the boiler circuitry
itself, so this is why I wish to infer what the boiler is doing from
the temp sensor graphs.

I could probably work out some algorithms myself, but if a tool (R lib/module)
already exists, I will use that.

Kind Regards


More information about the R-help mailing list