Architecture

InsideForest is organized into layers that separate rule extraction, region prioritization, and reporting.

Main layers

1. Forest layer: scikit-learn or PySpark forests managed by Trees.
2. Region layer: Regions merges rules, computes metrics, and creates ranges.
3. Label layer: Labels and the helpers in cluster_selector consolidate clusters.
4. Utility layer: Models and MetaExtractor document outputs.

Data flow

1. The user trains or supplies a forest.
2. Trees extracts rules and converts them into tabular structures.
3. Regions computes metrics, prioritizes, and builds regions.
4. cluster_selector helpers consolidate clusters based on the selected strategy.
5. Labels, Models, and MetaExtractor produce final reports.

Extensibility

• Add new backends by implementing the Trees interface.
• Define custom strategies by extending cluster_selector utilities.
• Extend reporting by adding helpers around Labels or MetaExtractor.

Continue with Configuration to learn how to tune parameters and with the API reference for detailed module docs.