Mixed_frequency data
#1
by
AdamBahelka
- opened
Hi, is there a way to use the model for a mixed frequency data? (eg. target is quarterly data while features are monthly?)
As far as i know there isnt. Whats the problem with aggregating the data before feeding it to the model? As long as ur input sequence has a higher frequency then the sequence u try to predict that shouldnt be a problem, or am i wrong?
Hi @AdamBahelka , the released Lag-Llama is a univariate model. So there's no "features" at all - you would just feed in the target data history and get predictions. And for now, only data of a single consistent frequency is supported.
arjunashok
changed discussion status to
closed