Anomaly Detection Archetypes and Markdown Summary

[JohT]

🚀 Feature

• Support fallback markdown includes for missing files: Now, markdown templates can specify fallbacks for includes that doesn't exist for conditional sections.
• Add sum of node embeddings contributing to the anomaly score (SHAP value): The new property anomalyNodeEmbeddingSHAPSum is now added to code unit nodes to reflect the sum of all absolute mean SHAP values that are related to node embedding dimensions. This shows how much all node embedding dimensions together influence the anomaly score.
• Output Markdown table with shapley explained top features: By executing the Python report generation there is now also a Markdown table included that shows the top features (globally) influencing anomaly detection and the anomaly score.
• 🌟 Add anomaly archetype labels and reports: Code unit nodes with outstanding traits now get these labels and properties:
  • Mark4TopAnomalyAuthority label and anomalyAuthorityRank property
  • Mark4TopAnomalyBottleneck label and anomalyBottleneckRank property
  • Mark4TopAnomalyBridge label and anomalyBridgeRank property
  • Mark4TopAnomalyHub label and anomalyHubRank property
  • Mark4TopAnomalyOutlier label and anomalyOutlierRank property
• Add betweenness centrality distribution: This additional plot shows how Betweenness Centrality is distributed across the code unit nodes (per type).
• 🌟Support Markdown reports: Markdown reports (scripts ending with Markdown.sh) are now supported additionally to the existing Python, Csv and Jupyter reports and will be picked up dynamically.
• 🌟 Add anomaly detection Markdown summary report: In the anomaly-detection report directory there is now one Markdown file called anomaly_detection_report.md that contains and summarizes all the single results and plots. It also contains a description on how the results were obtained and is readable for humans and large language models (context).

⚙️ Optimization

• Refine description and threshold variable names in anomaly detection queries.
• Refine description of the overridable environment variable MARKDOWN_SCRIPTS_DIR.
• Create sub directories for each anomaly detected code unit: Makes it much easier to navigate through the analysis results by having separate directories for e.g. Java_Package, Java_Type, ....

🛠 Fix

• Remove redundant markdown generation logging.
• Fix duplicate dash character in report artifact name