anomaly-detector

Anomaly detection identifies unusual patterns in your data using statistical methods, Isolation Forest, One-Class SVM, autoencoders, and Local Outlier Factor. Run the detector on transaction streams or system metrics to generate anomaly scores and flag outliers in real-time or batch workflows, enabling fraud detection and security monitoring.

Anomaly detection supports unsupervised, semi-supervised, and supervised approaches including Z-score, Isolation Forest, One-Class SVM, autoencoders, and LOF. These methods handle high-dimensional data, configure thresholds and anomaly scores, compute feature importances, and surface contributing factors to explain detected anomalies.

Yes. Anomaly detection processes transaction streams and log data in real-time or batch workflows, emitting anomaly scores into SpecWeave increment workflows for quick remediation. It covers fraud detection, intrusion detection, system health monitoring, and quality control with configurable thresholds.

No. Anomaly detection supports unsupervised methods (Z-score, Isolation Forest, LOF, autoencoders, One-Class SVM) that require no labels, plus semi-supervised and supervised approaches when labeled data is available. Choose the method that fits your data availability and use case.

Anomaly detection outputs anomaly scores, feature importances, and explanations of contributing factors. Results integrate seamlessly into SpecWeave increments, enabling downstream analysis, alerting, and automated remediation workflows.

Yes. Anomaly detection applies to transaction streams, system metrics, time-series data, and log data in both real-time and batch workflows. It scales to high-dimensional datasets and supports configurable thresholds for different detection scenarios.