Feature Selection and Generation¶

Feature Selection¶

Identify those variables or features that do not contribute to the classification task. In essence select d features out of available p features. This is also referred to as feature selection in the measurement space or feature selection.

Optimal methods¶

Optimal methods can lead to globally optimal solution but are computationally expensive -- exhaustive search; accelerated search; Monte Carlo methods.

Suboptimal methods¶

Suboptimal methods attempt to achieve a trade off between optimality and computational speed.

Filter Methods¶

Filter methods select features independent of the classification algorithm.

Wrapper methods¶

Wrapper methods select features based on the classification algorithm.

Feature Extraction¶

Transform from the p-dimensional feature space to a lower dimensional space. This is sometimes referred to as feature selection in the transformed domain or feature extraction.