skrobot

Add DatasetCalculationTask to wrap featuretools calculate_feature_matrix()
Change examples to use DatasetCalculationTask
Update documentation
Use categorical columns in pandas examples

• Fix problem of feature graph filenames (in case of Windows)
• Fix problem of long filenames in examples
• Upgrade to featuretools 0.23.0

Fix Read The Docs documentation build by adding .readthedocs.yml

Fix documentation, examples.

Add DeepFeatureSynthesisTask
Add EmailNotifier
All tasks can retrieve also data as dataframes (instead only as URL/file path)
Add examples with DeepFeatureSynthesisTask
Upgrade dependencies

Fix versionings in PyPI, Read the Docs, README.md, examples, etc

Also update all requirements.txt dependencies to latest.

Add Sphinx documentation for the source.

The PredictionTask returns in the output also the model's probabilities.

• The PredictionTask now works for probabilistic models and a threshold is provided (default value is 0.5)

• Metrics are now computed correctly

• Examples are re-generated

• The EvaluationCrossValidationTask now supports parameters threshold_selection_by and metric_greater_is_better for selecting the best threshold based either on a threshold or a metric

Update dependencies to latest version.

scikit-learn==0.23.1
joblib==0.16.0
numpy==1.19.1
pandas==1.0.5
numpyencoder==0.2.0
plotly==4.9.0
stringcase==1.2.0
scikit-plot==0.3.7
matplotlib==3.3.0

• PredictionTask class is added

• Source code file path is now optional and can be given with set_source_code_file_path() in Experiment class

• Accuracy metric is added in HyperParametersSearchCrossValidationTask class

• Changes in examples

This is the first release of skrobot.