Hi all, I am trying to use LightGBMCV to compute reliable cross validation for my ts model. I get an error and I think it may be due to gaps in some of my times series. Could you tell me if LightGBMCV can be used with gaps in time series data ? Thanks Fabrice.

It will run but the lags won't be correct if you have gaps

Thank you José, So I have the choice: _filling the gaps the best I can, _or making my own lagged features outside MlForecast am I right ?

Are your gaps regular? e.g. do you have data from monday to friday? or are they random? If they're random then I suggest filling them or generating the features in some other way, if they're regular you can use mlforecast

gaps are not regular in any one time series (I have thousands). Thank you for your answers and advice, I'll try that.