Hi Team,
I am creating segmented Time-series model with 220 combinations of time series(Customers-Product).
When I run this model on Data robot, It took more than 26 hours(1 day 2 hours).
Is there any option to have early stopping of these and get optimal results.
Please find the attached screenshot.
Thank you
Lokeshwaran
Here is the answer from my colleague
The answer is 220 segments multiply by ~20models for each segment divide by 4 workers = 1100 models per worker. 26hours/1100 models = les than 1.5 minutes per model, which is great result. Each segment is separate project with its own early stopping.
Recommendation might be to use Quick mode to reduce amount of models trained, or group segments to higher levels to produce less segments, or even try to do this without segmentation.