Abstract:This paper proposes a novel fast online methodology for outlier detection called the exception maximization outlier detection method(EMODM), which employs probabilistic models and statistical algorithms to detect abnormal patterns from the outputs of complex systems. The EMODM is based on a two-state Gaussian mixture model and demonstrates strong performance in probability anomaly detection working on real-time raw data rather than using special prior distribution information. We confirm this using the synthetic data from two numerical cases. For the real-world data, we have detected the short circuit pattern of the circuit system using EMODM by the current and voltage output of a three-phase inverter. The EMODM also found an abnormal period due to COVID-19 in the insured unemployment data of 53 regions in the United States from 2000 to 2024. The application of EMODM to these two real-life datasets demonstrated the effectiveness and accuracy of our algorithm.

What problem does this paper attempt to address?

This paper attempts to address the problem of detecting anomalous patterns in complex systems. Specifically, the paper proposes a new fast online anomaly detection method—Anomaly Maximization Anomaly Detection Method (EMODM), which detects anomalous patterns from the output of complex systems through probabilistic models and statistical algorithms. The paper particularly focuses on anomaly detection in real-time raw data without the need for special prior distribution information. ### Main Issues 1. **Anomaly Detection in Complex Systems**: How to efficiently and accurately detect anomalous states in complex systems. 2. **Real-time Detection**: How to quickly detect anomalous patterns in real-time data streams. 3. **No Prior Information Needed**: How to perform anomaly detection without special prior distribution information. ### Solution The paper proposes the Anomaly Maximization Anomaly Detection Method (EMODM), which is based on a two-state Gaussian Mixture Model (GMM) and can achieve real-time anomaly detection. The main features of EMODM include: - **Probabilistic Model**: Using a probabilistic model to detect anomalous patterns. - **Real-time Processing**: Capable of quickly detecting anomalies in real-time data streams. - **No Prior Information Needed**: Does not require special prior distribution information, making it suitable for various complex systems. ### Application Cases 1. **Short Circuit Detection in Circuit Systems**: Using the current and voltage output of a three-phase inverter, EMODM detects short circuit patterns in circuit systems. 2. **US Unemployment Insurance Data**: Using EMODM to detect anomalous periods in the unemployment insurance data of 53 regions in the US from 2000 to 2024, particularly the anomalous periods caused by COVID-19. ### Conclusion The application of EMODM on two real-world datasets demonstrates its effectiveness and accuracy, proving the method's capability in anomaly detection within complex systems.

Machine Learning for Complex Systems with Abnormal Pattern by Exception Maximization Outlier Detection Method

Anomaly Detection Algorithm Based on Electric Equipment

An Extreme Learning Machine for Unsupervised Online Anomaly Detection in Multivariate Time Series

Outlier detection based on improved SOM and its application in power system

Learning Multi-Pattern Normalities in the Frequency Domain for Efficient Time Series Anomaly Detection

Latent Laplacian Maximum Entropy Discrimination for Detection of High-Utility Anomalies

EAD: effortless anomalies detection, a deep learning based approach for detecting outliers in English textual data

Outlier Detection Method Based on Compressed Time Series and Voronoi Diagram for Power Data

Online Anomaly Detection Using Machine Learning and HPC for Power System Synchrophasor Measurements

Anomaly Detection Based on Machine Learning for the CMS Electromagnetic Calorimeter Online Data Quality Monitoring

Supervised Outlier Detection Based on Mahalanobis Ellipsoidal Learning Machine

A Deep Learning Approach to Anomaly Sequence Detection for High-Resolution Monitoring of Power Systems

Abnormal detection and early warning mechanism of power market transaction data: a practice and exploration

A Fast Outlier Detection Method for Big Data.

Detection and Classification of Anomalies in Power Distribution System Using Outlier Filtered Weighted Least Square

A novel energy-based online sequential extreme learning machine to detect anomalies over real-time data streams

EntropyStop: Unsupervised Deep Outlier Detection with Loss Entropy

New hybrid ensemble method for anomaly detection in data science

Outlier-Detection-Based EKF for Power Systems

Anomaly Detection for Power Consumption Patterns Based on Unsupervised Learning

Real-Time Outlier Detection with Dynamic Process Limits