Feature Selection for Mahout

This is a Java library for Apache Mahout for feature selection algorithms in MapReduce. So far the algorithma available are mRMR and Ranking.

Content

• src contains the source code of the library
• jar is the library you can use to run mRMR algorithm
• example provides a script with all the commands required to run mRMR

Unit Tests

Tests for the mRMR MapReduce job are not available, since mRMR is implemented taking advantage of the Distributed Cache, while MRUnit package does not provide any support for that.

Acknowledgement

This library has been developed as part of a project supported by Machine Learning Group and granted by Spinner