TSelect

Installation

Pip install

pip install tselect

Clone repository

(Temporary until published) Clone the repository.

git clone https://github.com/LorenNuyts/TSelect.git

Go to the newly created repository and install the requirements with pip.

pip install -r requirements.txt

Known issues

On Windows, the installation of the pycatch22 package can fail. Installing the package with the following command usually fixes this.

pip install pycatch22==0.4.2 --use-deprecated=legacy-resolver

Examples

Feature extraction with TSFuse

TSFuse requires the data to be in a specific format, see TSFuse for more information.

from tsfuse.construction.mlj20 import TSFuseExtractor
from tselect.channel_selectors.tselect import TSelect

[...] # load data, split in train and test set, etc.

extractor = TSFuseExtractor(transformers=tsfuse_transformers, compatible=compatible, random_state=SEED,
                                    series_filter=TSelect())
extractor.fit(x_train, y_train)
features_train = extractor.transform(x_train)
features_test = extractor.transform(x_test)

[...] # Training a model, etc.

Feature extraction with MiniRocket

from tselect.minirocket import MiniRocketExtractor
from tselect.channel_selectors.tselect import TSelect

[...] # load data, split in train and test set, etc.

# If fusion is used, the `views`, `add_tags`, and `compatible` arguments must also be specified for correct data transformation.
# More information on these arguments can be found on [TSFuse](https://github.com/arnedb/tsfuse#data-format).
extractor = MiniRocketExtractor(series_fusion=True, irrelevant_filter=True, redundant_filter=True)
extractor.fit(x_train, y_train)
features_train = extractor.transform(x_train)
features_test = extractor.transform(x_test)

[...] # Training a model, etc.

Feature extraction with Catch22

from tselect.catch22 import Catch22Extractor
from tselect.channel_selectors.TSelect import TSelect

[...] # load data, split in train and test set, etc.

# If fusion is used, the `views`, `add_tags`, and `compatible` arguments must also be specified for correct data transformation.
# More information on these arguments can be found on [TSFuse](https://github.com/arnedb/tsfuse#data-format).
extractor = Catch22Extractor(series_fusion=True, irrelevant_filter=True, redundant_filter=True)
extractor.fit(x_train, y_train)
features_train = extractor.transform(x_train)
features_test = extractor.transform(x_test)

[...] # Training a model, etc.

Fusion and filtering only

If you only want to use the fusion and/or filtering components of this package, you can use the FusionFilter class directly.

from tselect import FusionFilter

[...] # load data, split in train and test set, etc.

# Default arguments, please specify this for the used data
views = None
add_tags = id
compatible = lambda x: True

fusionfilter = FusionFilter(views=views, add_tags=add_tags, compatible=compatible)
x_pd_train = fusionfilter.fit(x_train, y, return_format='dataframe')
x_pd_test = fusionfilter.transform(x_test, return_format='dataframe')

[...] # Further processing of the fused and selected signals.