nicl-nno
commented
2 years ago Hello, Jacub!
The time series classification is not implemented yet, but it can be added to the framework with relatively minor changes.
If you want to make this change by yourself, start from the extending of DataTypesEnum class with 'ts_classification' option and adding of automatic feature extraction algorithms (e.g. quantile features, spectral features, etc) as data operation.
The modification of the pipeline optimizer is not required, since it is model-agnostic by design.
We are ready to help with any questions if necessary.
From the documentation and examples I see that both time series forecasting and tabular classification are possible with FEDOT. But is time series classification also possible? For example, automatic feature extraction to turn time series into tabular format with extracted features and further automatic classification?