Why does XGBoost predcontribs return one more value than the number of input features?
Classic MLLevel: IntermediateXGBoost · Prediction
Why does XGBoost pred_contribs return one more value than the number of input features?
Hints
- Local contributions need a reference value.
- Reconstruction includes that baseline.
Answer
The extra component is the bias/base contribution. Include it when reconstructing the margin from contributions, and preserve class/group axes for multiclass outputs; do not report it as an extra input feature.
Source and licence: XGBoost · Prediction · Apache-2.0 · answer: Quantoros editorial