From: plfriko@yahoo.de   
      
   Hi,   
      
   I'm using supervised machine learning to classify my data. The   
   approach I use as classifier is a decision tree (but could by any   
   other)- After constructing an appropriate decision tree, I would like   
   to measure the model's performance. What are standard measures in the   
   domain of statistics and artificial intelligence domain to estimate   
   performance of a classification algorithm?   
      
   So far, I've used a leave-one-out cross validation (due to the small   
   number of examples in the learning set which is about 400) to evaluate   
   the accuracy (classification error), i.e. how many examples in the test set   
   were incorrectly predicted. However, I don't think that this is sufficient   
   for a reliable performance evaluation. What else should I measure?   
      
   I'm not sure if a significance test would provide helpful information.   
   In my text book, they use the significance test to compare two   
   different classification algorithm w.r.t. to their absolute error   
   (they determine by a cross validation). Can a significance test be   
   also exploited to make performance assumption about a single classifier?   
   If so, what hypothesis should be tested?   
      
   Thank you.   
      
   Regards,   
   Tim   
      
   [ comp.ai is moderated ... your article may take a while to appear. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)