PERFORMANCE ANALYSIS

• DT system performs better than PCA system on training data but worse on test data

  
  

• DT system seems to memorize the training data and suffers with unseen data

  
  

• The trees built from four training sets are different, suggesting overtraining of the data

  
  

• DT system cannot differentiate between overlapping features

  
  

• Performance degrades if key features are left out but performance remains the same if extraneous features are left out

  
  

• DT system can detect features that give little discriminating information and discard them