We know that our time series pipelines require data with a regular frequency so that we can correctly specify the values of gap, max_delay, forecast_horizon and build features from these parameters.
Naturally, we need to ask how to handle irregularly spaced data. One thing we can try is sampling to a regular frequency in a component.
We know that our time series pipelines require data with a regular frequency so that we can correctly specify the values of
gap
,max_delay
,forecast_horizon
and build features from these parameters.Naturally, we need to ask how to handle irregularly spaced data. One thing we can try is sampling to a regular frequency in a component.
https://github.com/dataiku/dss-plugin-timeseries-forecast/blob/main/python-lib/timeseries_preparation/preparation.py