Anomalous time collection are a really critical enterprise.
If you concentrate on earthquakes, anomalies are the irregular seismic alerts of sudden spikes or drops in information that trace that one thing unhealthy is happening.
In monetary information, everybody remembers the Wall Road Crush in 1929, and that was a transparent instance of a sign with anomaly within the monetary area. In engineering, alerts with spikes can signify a mirrored image of an ultrasound to a wall or an individual.
All these tales stem from a really well-defined downside:
If I’ve a financial institution of regular alerts, and a brand new sign is available in, how can I detect if that sign is anomalous or not?
Be aware that this downside is barely completely different than the issue of detecting the anomaly in a given sign (which can also be a widely known downside to resolve). On this case, we assume that we get a complete new sign and we wish to know if the sign is sufficiently completely different than those which might be thought of “regular” in our datasets.