ML Signal Detection Pipeline

Autonomous machine learning for PV signal detection — feature extraction, random forest training, evaluation, and prediction on FAERS disproportionality data

5 toolsml-pipeline

Ml Feature Extract

Extract 12-element PV feature vector from FAERS 2x2 contingency table: PRR, ROR, IC, EBGM, log(cases), HCP ratio, consum

drugeventabcd

Ml Train

Train a random forest model on labeled PV feature data. Returns model_id, training accuracy, AUC, F1 score, and feature

samples

Ml Predict

Predict signal probability for drug-event pairs using a trained ML model. Returns 'signal' or 'noise' classification wit

model_idsamples

Ml Evaluate

Evaluate a trained ML model on held-out test data. Returns AUC-ROC, precision, recall, F1 score, accuracy, and confusion

model_idtest_samples

Ml Pipeline Run

Run the full autonomous ML pipeline end-to-end: ingest raw FAERS contingency data, extract 12 PV features, train random

datalabels