Prevent Score recipe from resampling in time-series forecasting
Hello! I have built a time-series forecasting model in the lab and am using it to forecast/score sales data. I have a long-format dataset with SKU codes, order dates and sales for each date. For each SKU, I have consecutive dates and no duplicates. Nevertheless, the score recipe keeps on resampling the data, which takes a very long time. How can I avoid this? I am using the NPTS model with weekly seasonality.
Can you help me? Thanks.
--
Best regards,
Xavier
Operating system used: Windows
Best Answer
-
Hi,
Unfortunately, the resampling step always happens but it won't change the data if all the time steps are already regular and equi-spaced.
Removing columns that are not used by the model might make the resampling faster. Other than that, partitioning the input dataset of the scoring recipe can make it faster because each partition will be scored in parallel.
