Icon

BDI_​Training

5 workflows in this folder and its subfolders

Icon
M12_​KNIME_​API

M11 KNIME serving workflow = REST endpoint equivalent of m11_app.py /predict. Container Input (JSON) -> JSON to Table -> rename hour to Hour -> Model Reader (pump_rf.model) -> Random Forest Predictor -> probability (4dp) + prediction (threshold 0.328) + risk_level + recommended_action (Thai) + model…

Icon
M12_​train_​save

M11 train-and-save: trains a KNIME Random Forest on the 5 pump features (temperature_c, temp_roll6_mean, temp_roll6_max, temp_delta, Hour) and writes the model to M11_KNIME_API/data/pump_rf.model via Model Writer. KNIME equivalent of training the model that m11_app.py serves.

Icon
M13_​pipeline

KNIME side-by-side equivalent of m9x_train.py: chronological split, past-only feature engineering, Random Forest (class-weight balanced analog), precision/recall + cost-aware threshold for imbalanced pump anomaly detection.