Wrapper Feature Set Evaluation

Contact: Martin Gütlein

Categories: Feature selection

Exposed methods:

Description:

The wrapper approach depends on the classifier that should be used with the resulting attribute subset.
Wrapper methods evaluate subsets by running the classifier on the training data, using only the attributes of
the subset. The better the classifier performs, usually based on cross-validation, the better is the selected
attribute set. One normally uses the classification-accuracy as the score for the subset. Though this technique
has a long history in pattern recognition, [JOH94] introduced the term wrapper that is now commonly used.

Background (publication date, popularity/level of familiarity, rationale of approach, further comments)
Standard feature selection method. Leads to superior results compared to Filter
methods. Slow. Resulting feature set is specific to the QSAR model that is used by the
wrapper.

Class-blind/class-sensitive feature selection
Class-sensitive feature selection

Filter/wrapper/hybrid approach
Wrapper

Type of Descriptor:

Interfaces:

Priority: Low

Development status:

Homepage:

Dependencies:
External components: WEKA

Technical details

Data: No

Software: Yes

Programming language(s): Java

Operating system(s): Linux, Win, Mac OS

Input format: Weka's ARFF format

Output format: Weka's ARFF format

License: GPL

References

References:
[JOH94] George H. John, Ron Kohavi and Karl Pfleger, Irrelevant Features and the Subset Selection Problem. ICML, 1994