Machine Learning has four general classes of forms: classification, predicting next value, exception detection, and discovering structure. Among them, Anomaly detection detects data points in data that does not fit well with the rest of the data. It has a wide range of using such as fraud detection, surveillance, diagnosis, data cleanup, and predictive maintenance.
Although it has remained considered in detail in academia, statements of anomaly detection have been limited to niche domains like banks, financial institutions, auditing, and therapeutic diagnosis, etc. However, with the advent of IoT, anomaly detection would possibly play a key role in IoT use cases such as monitoring and predictive maintenance.
Anomaly detection is a technique used to recognize unusual designs that do not agree to expected performance, called outliers. It has many applications in business, from interference detection (identifying strange patterns in network traffic that could signal a hack) to system health monitoring (spotting a malignant tumor in an MRI scan), and from fraud detection in balance card transactions to fault discovery in working environments.
This post investigates what is anomaly detection, different anomaly detection techniques, presents the key idea behind those techniques, and wraps up with a conversation on how to make use of those issues.
Point Anomalies. If an individual data example can be recognized as exceptional concerning the rest of the data (e.g. purchase with large transaction value)
Contextual Anomalies, If a data situation is anomalous in a particular context, but not otherwise ( anomaly if occur at a particular time or a certain region. e.g. large spike at the middle of the night)
Collective Anomalies. If several related data instances are anomalous concerning the entire data set, but not individual values. They have two variations.
Events in unexpected order ( ordered. e.g. breaking rhythm in ECG)
Unexpected value sequences ( underscored. e.g. buying a large number of expensive items)