How wueth/tree-like-pin works
Tree-Like-PIN is research code and data for a neural-network model for tabular prediction, especially insurance pricing. Its documented workflow is to model pairwise feature interactions with a shared feed-forward architecture, produce predictions, and support inspection of interaction effects and efficient SHAP analysis. The README also reports empirical comparison on French motor insurance data. No source files or implementation paths were provided, so the graph captures only documented capabilities and relationships; it does not infer code-level modules or execution details.
Components 9
PIN modeling
- Tree-like PIN
- Shared feed-forward network
- Pairwise interactions
- Predictions
Interpretation and evaluation
- Interaction inspection
- SHAP computation
- French motor data
- Benchmark models
Other components
- Model user
Connections 9
- Tree-like PIN → Shared feed-forward network: uses shared architecture
README.md - Shared feed-forward network → Pairwise interactions: captures interactions
README.md - Tree-like PIN → Predictions: predicts
README.md - Model user → Interaction inspection: inspects effects
README.md - Tree-like PIN → Interaction inspection: enables inspection
README.md - Tree-like PIN → SHAP computation: supports efficient analysis
README.md - Pairwise interactions → SHAP computation: limits computation to pairs
README.md - Tree-like PIN → French motor data: evaluated on
README.md - Tree-like PIN → Benchmark models: outperforms in reported accuracy
README.md