search

alkaline-ml / pmdarima

A statistical library designed to fill the void in Python's time series analysis capabilities, including the equivalent of R's auto.arima function.
https://www.alkaline-ml.com/pmdarima
MIT License
1.58k stars 232 forks source link

arima update takes forever #417

Open ThomasBourgeois opened 3 years ago

ThomasBourgeois commented 3 years ago

Describe the bug Updating an arima model seems to be very long. I'm adding only 36 new samples on an arimax with only one exogenous parameter, and the update takes nearly 10 seconds on my regular laptop. (I have to do this operation many many times, hence the issue on my side) I have tried to put max_iter to 0 or leave it as None, but this does not change the time taken to do the operation. Am I misunderstanding what update does, with max iter at 0 there is no fit operation, so I thought the update would be extremely rapid ?

Versions 1.8.0

padhariabhavesh commented 2 years ago

An ARIMA model with zero means is white noise, so it means that the errors are uncorrelated across time.

This doesn't imply anything about the size of the errors, so no in general it is not an indication of a good or bad fit.

ThomasBourgeois commented 2 years ago

An ARIMA model with zero means is white noise, so it means that the errors are uncorrelated across time.

This doesn't imply anything about the size of the errors, so no in general it is not an indication of a good or bad fit.

Is that an AI generated message ? Because that answer does not make any sense at all. It is not in connection with my message and what it says is nonsense : an ARIMA with zero means is not white noise, eg MA(1) has zero means and errors are clearly correlated hence not white noise.