features for multivariate time series, especially with mixture of categorical and continues values

[Sandy4321]

what is about features for multivariate time series, especially with mixture of categorical and continues values can you share some such a dataset (train and test ) with performance of your code

for example with multivariate time series with table per each label like target is YES date f1 f2 f3
dec 0.1 a 234 jan -0.5 a 456 feb 3.4 b 123 march 0.6 b 678

like target is NO date f1 f2 f3
dec -0.1 c 1234 jan 0.5 a 4456 feb 2.4 g 2123 march 1.6 b 6678

[nils-braun]

Hello @Sandy4321 In the moment tsfresh can only handle numerical values. Some of our feature extractors might also work well with categorical columns (such as everything related to counting values), but our full pipeline was really built for numerical values.

What you could try, is to transform the categorical values in numbers, and apply the feature extraction on them, but many extractors will be non-sense (what is the mean of that column?). It would be interesting to see anyways. I do not have any performance to show here - also because the real performance heavily depends on the ML method you are using after that.