What is it about?
It relies on the multivariate probability distribution associated with the observations. Rare events are present in the tails of the probability distributions that calls for using copula functions. The joint distribution of the random variables corresponding to the features is determined and we assign anomaly scores to the observations based on the density. It also localizes the subspaces where the observation has high anomaly score.
Featured Image
Photo by Will Myers on Unsplash
Why is it important?
There is huge amount of versatile data for performance monitoring. The signs of sub-optimal operation can remain hidden for a potentially long time. Many such hidden issues should be isolated and indicated to the network operator using a model-based anomaly detection and localization method.
Perspectives
The proposed method assigns high anomaly score to those observations that are found anomalous by alternative methods as well. It can provide information on the location of the anomaly. It can operate with missing data as well.
Roland Molontay
Budapesti Muszaki es Gazdasagtudomanyi Egyetem
Read the Original
This page is a summary of: Copula-Based Anomaly Scoring and Localization for Large-Scale, High-Dimensional Continuous Data, ACM Transactions on Intelligent Systems and Technology, June 2020, ACM (Association for Computing Machinery),
DOI: 10.1145/3372274.
You can read the full text:
Contributors
The following have contributed to this page
