ML Anomaly Detection Training: Start With the Base Rate

Most training on ML-based anomaly detection for cybersecurity stops at fitting the model. The half that decides whether a detection works is evaluation.