[source]

Recursive Feature Eliminator

Recursive Feature Eliminator (RFE) is a supervised feature selector that uses the importance scores returned by a learner implementing the RanksFeatures interface to recursively drop feature columns with the lowest importance until a terminating condition is met.

Note

The default feature ranking base learner is a fully-grown decision tree.

Interfaces: Transformer, Stateful, Verbose, Persistable

Data Type Compatibility: Depends on the base learner

Parameters

#	Name	Default	Type	Description
1	minFeatures		int	The minimum number of features to select from the dataset.
2	maxDroppedFeatures	3	int	The maximum number of features to drop from the dataset per iteration.
3	maxDroppedImportance	0.2	float	The maximum importance to drop from the dataset per iteration.
4	scorer	Auto	RanksFeatures	The base feature scorer.

Additional Methods

Return the final importances of the selected feature columns:

public importances() : ?array

Example

use Rubix\ML\Transformers\RecursiveFeatureEliminator;
use Rubix\ML\Classifiers\RandomForest;

$transformer = new RecursiveFeatureEliminator(30, 2, 0.05 new RandomForest());

References

---

1. I. Guyon et al. (2002). Gene Selection for Cancer Classification using Support Vector Machines. ↩

---

Last update: 2021-03-03