Scikit-disruption: machine learning toolkit for disruption analysis

Scikit-disruption is an open-source python library for disruption analysis. This library offers a set of tools for preprocessing training data, initiating machine learning models, and computing performance metrics that are relevant to disruption prediction applications. Scikit-disruption is designed to complement standard data science and machine learning libraries such as pandas, scikit-learn, and keras. It is designed to primarily interface with DisruptionPy [1,2], the data-processing framework which processes raw experimental and computational data into ML-ready datasets.

Scikit-disruption facilitates rapid prototyping of disruption prediction models. Using this new library, the PSFC Disruption Studies Group is actively exploring novel research areas, including but not limited to: new model architectures through semi-supervised learning and anomaly detection paradigms, disruption prediction during ramp up and ramp down phases, uncertainty quantification, robustness against sensor noise and error, and continual learning. The group also plans to consolidate previously developed architectures, such as the hybrid deep learner (HDL) [3], GPT-based classifier [3], and continual convolutional neural network (CCNN) [4], into this library and release pre-trained models as part of the application.