Ds-box

Added

- Add custom input unit and output unit in Dataoperations

Added

- Add GKEDataOperator for Airflow v1 and v2

Added

- Add DataInputBigQueryUnit and DataOutputBigQueryUnit

Added

- Add zero memory copy rolling functions (using NumPy)
- Add Outliers notebook example

Changed

- Remove timestamp examples in Shifter class doctests

Fixed

- Force pod name to maximum 64 characters in KubeDataOperator class

Added

- KerasApplicationFactory generating known NN architectures for image labelling (for instance Xception)
- Add inverse_transform method for TagEncoder
- Documentation for OPS part (Operators, Data Units)
- DataGlobalInputUnit and DataGlobalOutputUnit able to handle all dataframes APIs (Pandas, Vaex, Dask and co.)
- Tutorial notebook "Classic Auto-encoders architectures"
- Tutorial notebook "LSTM and CNN for text classification - Sentiment analysis applied to tweets in French"

Changed

- Update keras imports to tensorflow.keras
- Add OutlierMixin inheritance to outliers class
- Add a window parameter to MADOutliers estimator

Fixed

- Add __str__ method to multiple data units (cause crash in Apache Airflow if consulting task instance details)
- Update joblib import from sklearn.externals.joblib to joblib
- Re-aligned some pacakges importation to avoid future deprecation

Added

- Data Operator class to handle data operation in Apache Airflow
- Data Units which are used with Data Operators to manage data input and output streams
- DBConnector to connect RDBMS database
- PlasmaConnector for connection to Arrow Plasma store
- Calibration model techniques
- Some ensemble models (Rotation Forest)
- Model explanation methods (feature contribution and prediction interval in Random Forest)
- Categorical and time series pre-built feature engineering
- Feature selection methods (greedy)
- Simple markov chains, supervised and unsupervised Hidden Markov Models
- End-to-end NN architectures
- Outlier detections techniques (MAD, FFT, Gaussian Process, ...)
- NLP models
- Pre-built function to display common graphics (Confusion matrix, ROC curve, prediction interval viz)
- Metrics score
- Miscellaneous functions (object persistence, file name generator, ...)