Systems, Man, and Cybernetics

 

Home

• [P795] Correlation-based Feature Ranking for Online Classification
  Author(s): Osman Hassab Elgawi
  
  

  Abstract
  The contribution of this paper is two-fold. First, incremental feature selection based on correlation ranking (CR) is proposed for classification problems. Second, we develop online training mode using the random forests (RF) algorithm, then evaluate the performance of the combination based on the NIPS 2003 Feature Selection Challenge dataset. Results show that our approach achieves performance comparable to others batch learning algorithms, including RF.

  
   

Full Text (PDF)