how to use the perfcurve for ROC that has already been built
    5 visualizaciones (últimos 30 días)
  
       Mostrar comentarios más antiguos
    
Hi I have used the following listing of code to create an ROC curve based on some testing that I had carried out. I was wanting to use the perfcurve to get the AUC and some other statistics on the ROC curve. Is it possible to use the perfcurve on an ROC curve that has already been created or if not could someone provide me with help on how to use the perfcurve function on what I already have.
Kmax =max(max(Finalprobs')) %set Kmax
Kmin = 0;                    %set Kmin
K=linspace(Kmin,Kmax,1000) ;  %create 1000 K values evenly between Kmin and Kmax
Result1 =zeros(numel(K), 4); %create array "Result1"
for iK = 1:numel(K)          %start of for loop iK
    aK =K(iK);               %set aK to increment through the values of K
TP=0;
TN=0;
FP=0;
FN=0;       
      for confuseloopCO = 1:size(Finalprobs, 2)   %for loop for TN and FP
          if (Finalprobs(1,confuseloopCO)) > aK  
              TN=TN+1;                  
          else
              FP=FP+1;                  
          end
           if (Finalprobs(2,confuseloopCO)) > aK   %for loop for FN and TP
              FN=FN+1;                  
           else
              TP=TP+1;               
           end
      end
       Result1(iK,:) = [TP, FN, FP, TN]; %insert values of TP,FN,FP,TN into
                                         %Result1 during for loop iK
  end
sens = Result1(:,1)/1000
spec = Result1(:,4)/1000
x = (1-spec)
y = (sens)
figure 1;
plot(x,y,'r')
title('ROC Curve for QTc Data with Control Probability', 'FontSize', 16)
xlabel('1-Specificity', 'FontSize', 14)
ylabel('Sensitivity', 'FontSize', 14)
If needs be in can upload the entire code being used but it is rather long.
Cheers,
Ross
Further to this I have been able to use the trapz function to work out the area under the curve but would like to be able to perform the perfcurve on the data if at all possible.
0 comentarios
Respuestas (0)
Ver también
Categorías
				Más información sobre ROC - AUC en Help Center y File Exchange.
			
	Productos
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!