|
ppforest2 v0.1.0
Projection Pursuit Decision Trees and Random Forests
|
Grouped result of the variable importance measures. More...
#include <Evaluation.hpp>
Public Attributes | |
| types::FeatureVector | permuted |
| VI1 — per-variable permuted importance (forest only). | |
| types::FeatureVector | projections |
| VI2 — per-variable projection-coefficient importance. | |
| types::FeatureVector | scale |
| Per-variable σ used to rescale coefficients (columnwise sd). | |
| types::FeatureVector | weighted_projections |
| VI3 — per-variable weighted-projection importance (forest only). | |
Grouped result of the variable importance measures.
Produced by the variable_importance(...) overloads below. variable_importance(Tree) fills projections and scale; variable_importance(Forest) fills all four fields. Individual measures can be computed via vi_projections, vi_permuted, and vi_weighted_projections directly.
| types::FeatureVector ppforest2::VariableImportance::permuted |
VI1 — per-variable permuted importance (forest only).
| types::FeatureVector ppforest2::VariableImportance::projections |
VI2 — per-variable projection-coefficient importance.
| types::FeatureVector ppforest2::VariableImportance::scale |
Per-variable σ used to rescale coefficients (columnwise sd).
| types::FeatureVector ppforest2::VariableImportance::weighted_projections |
VI3 — per-variable weighted-projection importance (forest only).