FLEX-C

Forest of Local EXpert Classifiers (FLEX-C) is a semi-supervised framework for fault detection and identification designed to be robust when fault data is scarce and normal operation data is potentially contaminated.

Implementation Notice

FLEX-C models subclass the scikit-learn IsolationForest to inherit its optimized ensemble implementation. They can be used as a standard Isolation Forest with the usual fit and predict methods.

• Use inject_knowledge to supervisedly train the model after the unsupervised fit initialization.
• Use predict_labels to obtain a classification output.

A complete usage example is provided in the publication_experiments folder.