Cluster

Clustering Playbook

Prerequisites

• Session scratchpad: .mltoolkit/session.py
• Reference: {SKILL_DIR}/references/cluster_reference.py

Workflow

1. Read reference script.
2. Stage the reference bundle into .mltoolkit/: python {PLUGIN_ROOT}/scripts/stage_session.py --task cluster --dest .mltoolkit This copies cluster_reference.py as session.py, plus model_zoo.py and the _shared/ package, co-located.
3. Run EDA — elbow plot for choosing k: python .mltoolkit/session.py --data <DATA> --output-dir .mltoolkit --stage eda
4. Read elbow plot from .mltoolkit/artifacts/elbow.png — suggest a k to the user.
5. Run compare with chosen k: python .mltoolkit/session.py --data <DATA> --output-dir .mltoolkit --stage compare --n-clusters <K>