I have been trying tirelessly to run statsforecast and mlforecast on spark databricks but i have encountered several issues.
First, using statsforecast ensures my predictions for the future are always constant. E.g. prediction for next month and next 2 months are the same.
Second, using mlforecast sparkxgb or lgb does not work. There is always an error stating
Expecting property name enclosed in double quotes: line 1 column 5087 (char 5086)
Has anyone used statsforecast or mlforecast on Databricks (spark)? Do you have any pointers?
07/06/2023, 6:02 PM
I have been working on databricks and I encountered different issues with these packages, like error when importing. For you first issue, I think it might be the problem of your data or the model you choose because auto regressive models always tend to predict constant if there is not enough fluctuation.