Time Series Anomaly Detection for MATLAB
Interactively design and test anomaly detection algorithms for time series data.
The Time Series Anomaly Detection for MATLAB Support Package provides functions and an app for designing and testing anomaly detection algorithms to characterize normal behavior and detect anomalies in new data. You can detect a wide range of anomaly types, from simple outliers to complex multivariate patterns.
Apply anomaly detection to a variety of engineering applications such as improving flight test data, designing fault-tolerant control systems, matching patterns in digital health signals, analyzing industrial sensor data, and more.
Train and Test Anomaly Detectors
Use the Time Series Anomaly Detector App to interactively visualize data, train detectors, and compare statistical, machine learning and deep learning approaches. You can also train detectors at the command line using consistent interfaces. Test and evaluate detectors without requiring a large number of labeled anomalies.
Access Popular Ready-to-Train Detectors
Explore a curated library of ready-to-train anomaly detection algorithms commonly used in research and industry. Train preconfigured deep learning architectures like temporal convolutional networks (tcnAD), variational auto encoder LSTM hybrid network (vaelstmAD), and more. Explore machine learning algorithms like isolation forest and one-class SVM that train directly on raw time series data. Apply statistical process control detectors using standard control rules.
Generate Synthetic Anomalies
When you don’t have enough labeled anomalous data, inject a variety of rule-based synthetic anomalies into your time series data to test anomaly detectors and benchmark performance.
Identify Patterns in Time Series Data
Discover recurring or anomalous patterns without training a model. The Matrix Profile algorithm lets you search for recurring motifs and rare discords directly in your time series data—ideal for uncovering operational signatures, detecting unexpected behaviors, and accelerating root-cause analysis without the overhead of model development.
Evaluate Performance
Compute and visualize key performance metrics to understand detector performance, update detector parameters, and select the best algorithm for your data by comparing algorithms side by side.
MATLAB 版本兼容性
- 兼容 R2026a
平台兼容性
- Windows
- macOS (Apple 芯片)
- macOS (Intel)
- Linux
