A Review of Anomaly Detection Machine Learning and Deep Learning Techniques

In general, an anomaly may be defined as a variance or departure from what is expected or normal. According to historical records, the term “anomaly” was derived from the Greek word “anomalia,” which translates as “uneven” or “irregular.” Many examples of such abnormalities or deviations from normalcy have occurred in our everyday lives, and we have all seen them. For example, when a condition monitoring system detects any value or parameter of the machine that falls outside the minimum value limit to the maximum value limit, it beeps an alert. Similarly, when credit card fraud is detected, it notifies both the bank and the customer immediately. Currently, the most challenging problem is determining how to identify irregularities in data streams. An example of a data stream is a continuous stream of data that is continuously formed from any source and is referred to as streaming data (also known as streaming information). Finding anomalies in such a significant amount of data will be a time-consuming and challenging end eavour. In this paper, we will delve into the details of data streams and anomaly detection in data streams, examining a substantial number of papers and articles published on the topic in recent years.