Question about input_data of forecast_on_df method

[minelab-112stdnt-No7]

When using forecast_on_df, the input data is required to have three columns: unique_id, ds, and value_name. In the ds column, I tried storing data as integers representing the elapsed seconds or days, instead of using timestamp data of the datetime type.

However, this led to an error during the forecasting process, with the error message indicating that the ds column must contain timestamp data.

I would like to know if there is a way to make forecast_on_df work with data types other than timestamps in the ds column, or if I should consider using a different forecasting method.

[rajatsen91]

Hi, for your use case the forecast() method with freq=0 might be a better fit.