community
cancel
Showing results for 
Search instead for 
Did you mean: 

Alteryx Designer Ideas

Share your Designer product ideas - we're listening!

Rolling Average tool / enhance running sum tool

Would it be possible to add some additional options to the running total, in particular average (max, st dev. may be useful) to the running sum tool or create a rolling average tool where you can group by multiple fields and have a rolling window for the last x rows, which you select?  I know this can be done in the Multi-Row formula and with the moving summarise tool tool (http://www.chaosreignswithin.com/2014/12/moving-summarize.html), however the former is capped by 10,000 rows and alteryx crashes on my mac when trying to click on the multi-row formul tool with 6000 rows.  Also it takes a several hours to analyse 1 file, and would like to have a solution to do this on a daily basis. Also I need to be able to so a rolling average over the last 11,999 rows, as this would be the last 20 minutes of data using a 10Hz GPS unit. This would be a great tool I think.

 

 

2 Comments

As a quick test I tried sticking together a macro which might help. It can't do MAX but can do a rolling window Avg/StDev

 

https://gallery.alteryx.com/#!app/Rolling-Window-Count-Sum-Avg-StDev/57a7a2c4aa690a11a4c9d1c7

 

It only needs to do a Row-1 for the computation. Set the numerical input for the window size arbitarily at 100,000 but no reason not to increase as needed.

 

Should be a single pass but does do a join to get start of window for each point.

 

Max / Min had to do in a macro, might do an SDK version for fun which could do this

Alteryx Alumni (Retired)

I'd like to see a kernel-smoothing tool available in Summarize tool's add -> numeric menu.  The methods I'd like to see would include fixed- and variable-radius

 

  • unweighted (linear, uniform)
  • Gaussian
  • triangle
  • Epanechnikov

kernel smoothers, plus nearest-neighbor and perhaps local-regression smoothers.  See https://en.wikipedia.org/wiki/Kernel_smoother for details.  (For the uninitiated, fixed-radius unweighted kernel smoothing is what e.g. Excel calls a "moving average.") 

 

Kernel smoothing is possible in multiple dimensions.  When the number of dimensions exceeds three, the method's value usually degrades rapidly, so it'd suffice to support up to three dimensions.