Analysis of Deep Anomaly Detection Algorithms

Citations Over Time

Abstract

Anomaly detection, or outlier detection, is an important branch of machine learning. It has a wide range of applications, such as industrial damage detection, fraud detection and video surveillance detection. However, there are many challenges in anomaly detection. For example, it is difficult to define whether it is normal or abnormal as the anomalies are different in different scenarios, and the normal behavior is constantly changing. There are several methods of anomaly detection, including graphic method, statistical method, distance based method, density based method and model-based method. This paper mainly introduces the distance based method k-th nearest neighbor (KNN), density based method local outlier factor (LOF), model-based method isolation forest (iForest), and the author intends to find out their advantages and disadvantages. Finally, it is found that different algorithms work well in their own area. In fact, most of the algorithms has the problem of the curse of dimensionality. Additionally, the author points out that anomaly detection, which is based on semi supervised or weak supervised, is likely to be the future research direction.

Related Papers

• → Why is this an anomaly? Explaining anomalies using sequential explanations(2021)26 cited
• → Online Detection of Anomalous Sub-trajectories: A Sliding Window Approach Based on Conformal Anomaly Detection and Local Outlier Factor(2012)11 cited
• → Identification of outliers in pollution concentration levels using anomaly detection(2016)6 cited
• → Efficient Algorithms for Local Density Based Anomaly Detection(2017)4 cited
• Improving Influenced Outlierness(INFLO) Outlier Detection Method(2013)