What Is Anomaly Detection? Know The Different Techniques For It!
Before going into anomaly detection, a person should know what is the anomaly and the various types. Any deviation from the normal course that could lead to data loss or the items or events that are not in the expected pattern is known as the anomaly. The detection of such items and events is known as Anomaly Detection. The anomalies could be of three types, namely point, contextual and collective.
The point anomaly is one where only one data point is far from the rest. The contextual anomaly refers to a content-specific abnormality. A collective anomaly is detected by analyzing a set of instances showing variance.
Here are some techniques for the same, let us go through them quickly:
- Simple Statistical Methods: These methods use various statistical tools like mean, mode, median, and quantiles. The deviation from the mean signifies a variation and a rolling window is needed to compute the averages.
- Clustering based anomalies: This is one of the most common yet popular subjects in anomaly detection. It simply specifies that similar data points tend to belong to the similar cluster and the distance from the centroids define the anomaly.
- Density Based anomaly detection: This technique is based on the algorithm of k-based nearest neighbors. It works on an assumption that dense neighborhood is far away. The nearest set of data points are evaluated.
So, these are the most popular techniques used for the detection of anomalies. They should be detected in time using various techniques of Anomaly Detection and proper measures should be taken to address the problem.
Comments
Post a Comment