This is a programming/algorithm question, but mathematical insights would probably be very handy.
In essence I have 2000 arrays of electrical power output data, each consisting of around 20-40 positive integer entries. Most often they are in a certain bandwidth, say between 20,000 and 30,000:
[20131 22302 24300 .....]
[27000 28030 300000 .....]
each array is a power output spike where each number is milliwatts and represents a data at a certain 'second in time', so 10 entries would mean 10 seconds of measurements.
What I am interested in is somehow clustering these arrays by a computer program. Some of them are in the same sort of range and it would be nice to recognise patterns, where i can 'bucket' the arrays and then eventually label them as belonging to the same category.
I think this can potentially be solved by unsupervised machine learning algorithms, but i wonder if that would be the best way.
I am a programmer, and I could come up with some possible solutions, but I feel those might be a bit naive.
Any pointers/approaches/strategies would be appreciated!