Is there a way to create a blank TimeSeries object that you can append (concatenate or stack) to similar to list appending? about darts

[madtoinou]

Hi @gvas7,

I am assuming that the covariates you want to include in your optuna trial are actually columns/components of a single multivariate TimeSeries (hence, a unique and shared time index) and not distinct Timeseries with different time indexes.

You can simply use the __getitem__() operator to extracts the components you want to use for a given trial with:

# ts_covs is the master TimeSeries containing all the covariates, select_covariates is a list of strings
future_covariates = ts_covs[selected_covariates]

If the covariates are stored in separate ts with exactly the same time index and you want to use the stack method, you can use the snippet below. However, this is approach is not recommended because you will recreate a new TimeSeries at each iteration.

trial_ts = None
# select_covariates is a list of univariate TimeSeries where all the TimeSeries have the exact same time index
for cov in selected_covariates:
    if trial_ts is None:
        trial_ts = cov
    else:
        trial_ts = trial_ts.stack(cov)

[gvas7]

So simple and I overthought it. Thank you for taking time to explain this and the first example is what I wanted to do and does exactly what I wanted. As your package gets more popular, more people like me will show up asking questions like this one :(