Our application area of machine learning is in outlier detection, where the aim is to find instances that do not obey the general rule. The typical instances share characteristics that can be simply stated, and instances that do not have them are an anomaly. This model covers the typical instances and then any instance that falls outside is an exception. An outlier may indicate an abnormal behavior of the system or may indicate an incident requiring attention.
Outliers may also be recording errors (for example, due to faulty sensors) that should be detected and reviewed to get reliable statistics. An outlier may also be a novel, previously unseen but valid case, which is where the related term, novelty detection, comes into play.