Recursive Feature Selection for Enhancing Genetic Disease Prediction

From OpenCog
Jump to: navigation, search

At a Glance

  • Code: http://code.google.com/p/openbiomind/
  • License: GPLv2
  • Language: Java
  • Description: Implement recursive feature elimination(RFE) for OpenBioMind so that users of the gene microarray analysis software could obtain the top genes that contribute to a disease.

Major Project Milestone

  • Implement Recursive Feature elimination
  • Incorporate CFS FeatureRanking
  • Incorporate SVM FeatureRanking

Project Timeline

Week of: 2008-06-01
  • Familiarized with the class hierarchy of OpenBioMind.
  • Ran test-cases included in OpenBioMind's demonstration datasets.
  • Sketched the rough RFE architecture before coding.
Week of: 2008-06-08
  • Implemented GPRanking class, that ranks features based a pre-determined GP trials.
  • Implemented RecursiveFeatureElim class, that serves as the runner for the Recursive Feature Elimination.
TODO: 2008-06-08
  • Figure out how to maximize the performance of RFE, as RFE in current test-cases are quite stochastic. Should RecursiveFeatureElim backtrack or continue more trials of reducing feature-set even if the performance has declined?
  • How do we optimize the number of tasks ran?
  • Comment all code written so far and make sure that newly implemented classes conform to existing architecture and style.