[TriPython] Time Series Matching

[Will Spearman]

Thanks for the suggestions so far! I have not fully read the links yet, but
I will. Let me clarify the situation if that will help.

The data is from operating system performance counters collected from
within a single system. For example, series A is operations per second and
series B is latency. Both series are sampled every 1 second, but possibly
with offsets from each other, and possibly with missing samples. The
problem I am trying to solve is that I only want to see the calculated
latency per op for sample periods where both series had samples and those
samples' timestamps were less than 1 second apart.

On Fri, May 29, 2015 at 10:58 PM Kurt Grandis <kgrandis at gmail.com> wrote:

> I'm not quite sure I follow the criteria for sameness or elimination
> between the time series. Are each series independent sensor readings for
> the same environment?
>
> "Close" could mean a lot of things depending on how you're looking at the
> data--what dimensions or time frames you're considering. Do you have an
> idea of what the variance about each reading might be? Something so you can
> model "closeness" around each point.
>
> It sounds like you might want to explore transforming these readings into a
> continuous representation maybe using the above distribution model about
> each event. You can play with things like convolving the signals or using
> something like lowpass filters to trim extraneous readings.
>
> If you're looking to align the signals you can look into something like
> cross-correlation. It sounds like another approach depending on how many
> errant events there are is find the max cross correlation and remove the
> timestamps without partners; repeat.
>
> Not sure of the nature of the data you're dealing with, but hope something
> above helps.
> _______________________________________________
> TriZPUG mailing list
> TriZPUG at python.org
> https://mail.python.org/mailman/listinfo/trizpug
> http://tripython.org is the Triangle Python Users Group
>