Glossary · Phase 3 · Machine learning
Anomaly detection
Finding the points that don't fit the pattern of the rest, usually without labels, and ranking them for people to review.
Try it now
Each dot is a payment scored by a detector. Lower the share that's fraud, then drag the threshold all the way up.
Learn it properly
Anomaly detection · Masking, robust scores with the median and MAD, baselines from the past, isolation forests and thresholds from a review budget. About 35 minutes.
Other terms from the same lesson
- isolation forest: An anomaly detector that splits the data at random; points isolated in only a few splits are scored as unusual.
- median absolute deviation: The median of each value's distance from the median: a measure of spread that a few extreme values can't drag.
Never trained an AI?
Teach a small one to write, right in your browser. It takes about 20 seconds and needs no account.