Définition
Anomaly detection helps you find something that is rare but could be important: fraud, hardware failure, data error, suspicious login, unusual expense, or new type of user behavior. Often the task is difficult because there are few examples of anomalies or they are unknown in advance.
Exemple
The payment system notices that a customer suddenly makes a series of unusual transactions from another country and flags the event for review.
Pourquoi c'est important
The term is important for security, finance, industry, analytics and monitoring: a rare event can be more valuable than an average trend.
Fonctionnement
The model learns the normal behavior or distribution of the data and then evaluates how much a new example deviates from what is expected.
Où c'est utilisé
- antifraud
- equipment monitoring
- finding errors in data
Limites
Not every unusual event is a problem. Systems often produce false positives, so a threshold, context, and human verification are needed.
FAQ
Why is “Anomaly Detection” useful to know?
The term is important for security, finance, industry, analytics and monitoring: a rare event can be more valuable than an average trend.