1

I'm trying to detect if there's any pattern (seasonality) between login attempts on a machine using python.
After researching a little bit i stumbled upon the library pmdarima, providing statistical tests such as the Canova-Hansen test to estimate the seasonal differencing term.

All the tests however requires knowing apriori the parameter m which indicates the number of observations per seasonal cycle.
However data are observed all the time every millisecond therefore one day i could have 500 logins, another day i could have 70 logins and another one none at all.

let's say i want to test seasonality in one month period, or one week period,how should i go about selecting the coefficient for the test?
Thanks in advance

nonoDa
  • 111
  • 2
  • @RichardHardy unfortunatly it could be useful for R but i think a python implementation could be troublesome as pointed out here https://stackoverflow.com/questions/61464805/findfrequency-spec-ar-equivalent-in-python – nonoDa Jan 05 '21 at 08:02
  • While implementation is mostly off topic here, hopefully the statistical reasoning in the linked thread addresses your question. – Richard Hardy Jan 05 '21 at 08:06

0 Answers0