What Is Feature selection?

Feature selection is the process of choosing a subset of available input variables for a model. Methods include filters based on association with the target, wrapper methods that compare subsets through model performance, and embedded methods that select or shrink variables during fitting. Its aims can include better generalisation, lower complexity, lower cost, and greater interpretability.

In time-ordered financial data, feature selection is part of the model-training procedure. It must be performed separately using each fold’s training data; ranking features using validation or test rows leaks outcome information and makes subsequent performance estimates optimistic. A selected feature set should be checked for stability across folds and with feature ablation, because a high model-importance ranking does not by itself establish incremental value.

See also

• Variable selection

• Feature generation

• Feature ablation

Research and literature

• Isabelle Guyon and André Elisseeff, An Introduction to Variable and Feature Selection, Journal of Machine Learning Research 3, 2003.