I am looking for some suggestions on what techniques to use for stock data. Specifically i am looking for some methods to both de-noise data, and then find ways to transform the smoother data into a zero-centered format (i believe stock data is referred to as a non-stationary signal, is that correct?). I realize that to zero center my data, i'm probably going to want to look at chunks of the data and work on those individually. Was wondering if there was a standard or recommended technique for this. From there it should be far easier to pick the peaks and valleys i want.

I'm planning on using optimization methods to find the peaks and valleys i find, so the method doesn't need to be static but hopefully will have a few params i can play around with to determine the granularity and scope of the peak detection.

  • $\begingroup$ Hi: It's generally not a good idea to look at prices when analyzing stock data because they are excessively non-stationary. Returns (either first difference of log prices or arithmetic returns if horizons are long ) are what are used in practice. Obviously, returns can be non-stationary also but not nearly as poorly behaving as prices. Think of it this way. A 100 dollar stock goes up one 1 dollar and a a 10 dollar stock goes up 1 dollar. Which one do you prefer to be holding beforehand if you had 100 dollars to spend on stocks ? $\endgroup$ – mark leeds Oct 6 '19 at 5:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.