Problems with sample_period>30min in disaggregation

I have been working with the REDD dataset and with a dataset created by me (with a sample_period=1s in all meters). In both cases, when I try to disaggregate (CO and FHMM) the mains with a sample_period greater than 30 minutes, no process is performed, and the output.h5 file is empty.

This function I am using without problems for periods less than 30 minutes: disaggregate(test_elec.mains(), output, sample_period=3600)

Does anyone know why this happens?

Thanks

[PMeira]

I believe the main issue is the check for the minimum number of samples (defaults to 100). Try something like

#...
clf = CombinatorialOptimisation()
clf.MIN_CHUNK_LENGTH = 1
sample_period = 3600
clf.train(top_5_train_elec, sample_period=sample_period)
clf.disaggregate(test_elec.mains(), output, sample_period=sample_period)

There could be warnings about number of clusters (with such a low sample rate, sometimes there is only one, that is, no events for some appliances), otherwise it seems to work fine. Using .disaggregate_chunk() and iterating the chunks much like in the old example notebooks also seems to work as expected.

[PMeira]

Mathematically, I don’t see any inherit limitation for the methods. I don’t recall anything else related to the sample period or resampling that limits the upper range of the sample period, but it’s a possibility.

I’ll try to check this on the weekend. Should be easy to check. If you want to give it a go, try stepping through the code with a debugger.