From: p.rijnbeek@erasmusmc.nl   
      
   Can anyone help me with the following:   
      
   In non-fuzzy classification systems the performance measures often used to   
   compare systems are the accuracy, sensitivity and specificity as you all   
   know. What kind of performance measures are used in the fuzzy environment?   
   The performance measures above can only be used if the results are   
   dichotimized to correct and incorrect cases. At the end of the inference in   
   a fuzzy system you have to decide if the outcome is correct or not so you   
   have to dichotimize the defuzzified outcome by saying > 0.5 is correct lower   
   is incorrect? Isn't that a waist of information, e.g,   
      
   we have two classes and after the defuzzification the following values are   
   found:   
      
   u(A)=0.4   
   u(B)=0.6   
      
   suppose the correct class is A than the outcome is not correct when you use   
   a threshold of 0.5 but you can say that it did find some indications for   
   class A. How do you account for that in the performance calculations of the   
   classifier??   
      
   Do you calculate the absolute difference between the outcome and correct   
   membership of the three classes is that the solution and is that the   
   performance measure used in most published articles (no one tells that   
   explicitly?):   
      
   say u(A)=1, u(B)=0 is the correct outcome than this case gives an error of:   
      
   |1-0.4|=0.6 which is summed for all casses and divided by the number of   
   cases is that the trick?   
      
   The big problem is that, OK with fuzzy logic you can use all fuzziness of   
   the problem in the calculations, but in the end you have to make a decision   
   what the class is so you have to make a non-fuzzy decision which overrules   
   all advantages and only puts the crisp threshold in the end!!??   
      
   I hope someone can help me with this.   
      
   Thanks   
      
   Peter Rijnbeek   
   Erasmus Medical Center Rotterdam   
   The Netherlands   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)