Skip to content
BENIBIX Learn

Anomaly Detection

Also known as Outlier Detection, Novelty Detection

Finding the rare cases that do not fit, like a fake charge on a bank card.

Think of it like

Think of it like a drawer of white socks. One red sock stands out.

Example

Jack's card is normally used for small shop payments in London. One night it buys 3,000 euros of phones abroad. The bank's model flags it at once.

Why it matters

The rare cases are often the costly ones. Catching them early saves money and sometimes lives.

Definition

Anomaly detection is the task of finding the rare cases that do not fit the normal pattern. The model learns what normal looks like from lots of everyday data. Then it flags anything far from normal. It is used to catch card fraud, broken machines and break-ins. There are often very few bad cases to learn from, so it learns normal instead.
The line stays low for weeks. The one spike is far from normal, so it gets flagged.

Common mistake

Flagging everything that is just new. A first trip abroad is odd, but it is not fraud. Too many alarms get ignored.

Key takeaways

  1. 1
    Learn normal. From lots of everyday data.
  2. 2
    Flag the odd. Far from the usual pattern.
  3. 3
    Rare but costly. Fraud, faults and break-ins.

Learn it properly