17
Wrappers for feature subset selection Amir Razmjou

Wrapper feature selection method

Embed Size (px)

Citation preview

Page 1: Wrapper feature selection method

Wrappers for feature subset selection

Amir Razmjou

Page 2: Wrapper feature selection method

Benefits of Feature Subset Selection

• Too many dimensions; Elimination of the curse of dimensionality problem

• Improved model and classifier performance• Simple Models and elimination of over-

fitting• Faster training times.

Page 3: Wrapper feature selection method

The Process

Page 4: Wrapper feature selection method

Feature Selection Methods

Wrapperuse a search algorithm to search through the

space of possible features and evaluate each subset by running a model on the subset

Risk of over fitting to the modelComputationally expensive

EmbeddedEmbedded in and specific to a model

FilterSimilar to Wrappers in the search approachSimpler filter is evaluated

Page 5: Wrapper feature selection method

Hypothetical Concept

Page 6: Wrapper feature selection method

Definition 2

Page 7: Wrapper feature selection method

Contradiction 2

Page 8: Wrapper feature selection method

Definition 3

Page 9: Wrapper feature selection method

Contradiction 3 All features are irrelevant

Page 10: Wrapper feature selection method

Definition 4

Page 11: Wrapper feature selection method

Contradiction 4every feature is relevant

Page 12: Wrapper feature selection method

Definition 5 – Strong Relevance

Page 13: Wrapper feature selection method

Definition 6 – Weak Relevance

Page 14: Wrapper feature selection method

Conclusion 1

Page 15: Wrapper feature selection method

• Forward selection methods: these methods start with one or a few features selected according to a method specific selection criteria. More features are iteratively added until a stopping criterion is met.

• Backward elimination methods: methods of this type start with all features and iteratively remove one feature or bunches of features.

Page 16: Wrapper feature selection method

Relief

• Evaluates the worth of an attribute by repeatedly sampling an instance and considering the value of the given attribute for the nearest instance of the same and different class. Can operate on both discrete and continuous class data.

Page 17: Wrapper feature selection method

Relief

• Relief does not help with redundant features. If most of the given features are relevant to the concept, it would select most of them even though only a fraction are necessary for concept description