Efficient Generation and Selection of Combined Features for Improved Classification

  • Ahmad N. Shono

Student thesis: Master's Thesis


This study contributes a methodology and associated toolkit developed to allow users to experiment with the use of combined features in classification problems. Methods are provided for efficiently generating combined features from an original feature set, for efficiently selecting the most discriminating of these generated combined features, and for efficiently performing a preliminary comparison of the classification results when using the original features exclusively against the results when using the selected combined features. The potential benefit of considering combined features in classification problems is demonstrated by applying the developed methodology and toolkit to three sample data sets where the discovery of combined features containing new discriminating information led to improved classification results.
Date of AwardMay 2014
Original languageEnglish (US)
Awarding Institution
  • Computer, Electrical and Mathematical Sciences and Engineering
SupervisorVladimir Bajic (Supervisor)


  • Classification
  • Combined
  • Discovery
  • DCFD
  • Features

Cite this