HGAD (Heterogeneous Graph based Anomaly Detection) for Cloud Systems

HGAD is a Python-based research project that implements a heterogeneous graph anomaly detection framework for cloud system monitoring. The project uses PyTorch and PyTorch Geometric (PyG) to process the Server Machine Dataset (SMD) and evaluate anomaly detection performance.

Project Structure

• data/ — dataset loading, preprocessing, and raw data storage
• model/ — model implementations, including heterogeneous graph networks
• hypergraph/ — graph construction and correlation utilities
• train/ — training scripts and detector training logic
• results/ — experiment outputs, logs, and evaluation summaries
• docs/ — academic documentation, thesis drafts, meeting notes, reports, and changelogs

Documentation

The docs/ directory contains all thesis-related documentation for the HGAD project:

• docs/README.md explains the documentation directory structure and naming conventions
• docs/thesis/ contains thesis drafts and changelogs
• docs/reports/ contains interim and progress reports
• docs/meetings/ contains meeting minutes and action item records

For details on document layout and versioning, see docs/README.md.