sir, following is the error in my code while calculating ROC curve

Error using main_new (line 131)
You must pass scores as a vector of floating-point values.

%%Dorsal hand vein recognition using  SVM

clc
clear all;
c=[];
addpath train;
addpath test;
mapping=getmapping(8,'u2');                                 %LBP

         W=[0,0,0,0,0,0,0; ...
            0,1,1,1,1,1,0; ...
            0,1,2,4,2,1,0; ...
            0,1,4,4,4,1,0; ...
            0,1,2,4,2,1,0; ...
            0,1,1,1,1,1,0; ...
            0,0,0,0,0,0,0];

   %%Training images

   for folder_idx = 1 : 20             % no of classes
   for i = 1 : 9                             % no of images per class

   thisfile = fullfile('train', num2str(folder_idx), [num2str(i) '.bmp ']); 
   B = imread(thisfile ); 
   X = double(B);
    X = imresize(X,[60 60],'bilinear');  
    H2=DSLBP(X,mapping,W);
    Gray=X;
    Gray=(Gray-mean(Gray(:)))/std(Gray(:))*20+60;
    lpqhist=lpq(Gray,3,1,1,'nh');       
    % imshow(lpqhist);
    a=[H2,lpqhist];
    c=[c;a];
    disp(sprintf('Done',i));

   end 
   end

%%Testing images

d=[];

for folder_idx = 1 : 20     %no of classes
 for i = 1 : 3                     % no of images per class

   thisfile = fullfile('test', num2str(folder_idx), [num2str(i) '.bmp ']); 
   B = imread(thisfile ); 
    X = double(B);
    X = imresize(X,[60 60],'bilinear');  
    H2=DSLBP(X,mapping,W);
    Gray=X;
    Gray=(Gray-mean(Gray(:)))/std(Gray(:))*20+60;
    lpqhist=lpq(Gray,3,1,1,'nh');       
    a=[H2,lpqhist];
    d=[d;a];
    disp(sprintf('Done',i));
   end 
   end
P_train=c;
P_test=d;

% %%PCA low dimension reduction
% 
P_train = P_train';

% if classes are 20 then eiganvectors not exceed then 179 
model = perform_pca(P_train,179);                       %rank(P_train)-1

test_features= linear_subspace_projection(P_test, model, 1);        
P_train=model.train';
P_test=test_features';

%%Normalisation

    P_train=P_train/256;
    P_test=P_test/256;
% %%%%%%%%load label %%%%%%%%%%%%
 train_label=load('train_label.txt');  
 test_label=load('test_label.txt');    

% P_train = P_train';
% P_test = P_test';

%%classification K Nearest Neighour

% fit a kNN classification model
  mdl = fitcknn(P_train,train_label,'Distance','euclidean','NumNeighbors',1);    %euclidean

%  model=svmtrain(train_label,P_train);
% [predicted_label, accuracy, decision_values]=predict(test_label,P_test,model);

% apply the model to test data and compute test error
     [predictlabel,score,cost] = predict(mdl,P_test);

%[predictlabel, score] = resubPredict(mdl,P_test);
%  testError1 = sum(test_label ~= predict_label)/numel(test_label);

   % Fit probabilities for scores

      %groundTruthNumericalLable          = [ones(20,1); zeros(20,1); -1.*ones(20,1)];
      %[FPR, TPR, Thr, AUC, OPTROCPT]  = perfcurve(predictlabel(:,1), score(:,1),'Good');
  %   [~, Accuracy, Thr]  = perfcurve(predictlabel(:,1), score(:,1), 1,'yCrit','accu');
  %    figure,
  %     plot(Thr,Accuracy,'r-');
  %     hold on;
  %     plot(Thr,Accuracy,'bo');
  %     xlabel('Threshold for ''good'' Returns');
  %     ylabel('Classification Accuracy');
  %     grid on;
  %     
  %     disp('Recognition rate');
  %     [maxInd] = max(Accuracy)                  %display max index
  %     disp('threshold value');
  %     Thr(maxInd)                                     %displays threshold value

score = double(score);

[tpr, fpr, thresh] = roc(predictlabel,score); % The ROC function

[fpr2, tpr2] = perfcurve(predictlabel,score,1); % The perfcurve funtion

plotroc(predictlabel,score); % calculating and plotting altogether

figure(); plot(fpr, tpr); title('ROC'); xlabel('False Positive Rate'); ylabel('True Positive Rate'); % Plotting the ROC results

figure(); plot(fpr2, tpr2); title('perfcurve'); xlabel('False Positive Rate'); ylabel('True Positive Rate'); % Plotting the perfcurve results

Answer 1

[Walter Roberson]

score returned by predict is an samples x classes array giving probabilities for each class. https://www.mathworks.com/help/stats/classificationknn.predict.html

perfcurve requires a vector of values for the scores https://www.mathworks.com/help/stats/perfcurve.html#bunsogv-scores .

[Balaji M. Sontakke]

sir, I got a vector, score from this functions output [predictlabel,score,cost] = predict(mdl,P_test); but that score vector contains only 0 and 1 of size 60 X 20. I provide this score into roc function but this score and score from predict is same or not? how I can obtained score vector which I want to provide in roc function

[Walter Roberson]

Read https://www.mathworks.com/help/stats/perfcurve.html#bupy9b3-1

[Balaji M. Sontakke]

sir, I read a link and did accordingly, but results are like recognition rate 100 at threshold 1 and curve attach here, is it correct?

%% Dorsal hand vein recognition using SVM

clc clear all; c=[]; addpath train; addpath test; mapping=getmapping(8,'u2'); %LBP

       W=[0,0,0,0,0,0,0; ...
          0,1,1,1,1,1,0; ...
          0,1,2,4,2,1,0; ...
          0,1,4,4,4,1,0; ...
          0,1,2,4,2,1,0; ...
          0,1,1,1,1,1,0; ...
          0,0,0,0,0,0,0];

 %%Training images

 for folder_idx = 1 : 20             % no of classes
 for i = 1 : 9                             % no of images per class

 thisfile = fullfile('train', num2str(folder_idx), [num2str(i) '.bmp ']); 
 B = imread(thisfile ); 
 X = double(B);
  X = imresize(X,[60 60],'bilinear');  
  H2=DSLBP(X,mapping,W);
  Gray=X;
  Gray=(Gray-mean(Gray(:)))/std(Gray(:))*20+60;
  lpqhist=lpq(Gray,3,1,1,'nh');       
  % imshow(lpqhist);
  a=[H2,lpqhist];
  c=[c;a];
  disp(sprintf('Done',i));

 end 
 end

%% Testing images

d=[];

for folder_idx = 1 : 20 %no of classes for i = 1 : 3 % no of images per class

 thisfile = fullfile('test', num2str(folder_idx), [num2str(i) '.bmp ']); 
 B = imread(thisfile ); 
  X = double(B);
  X = imresize(X,[60 60],'bilinear');  
  H2=DSLBP(X,mapping,W);
  Gray=X;
  Gray=(Gray-mean(Gray(:)))/std(Gray(:))*20+60;
  lpqhist=lpq(Gray,3,1,1,'nh');       
  a=[H2,lpqhist];
  d=[d;a];
  disp(sprintf('Done',i));
 end 
 end

%% P_train=c; P_test=d;

% %% PCA low dimension reduction % P_train = P_train';

% if classes are 20 then eiganvectors not exceed then 179 model = perform_pca(P_train,179); %rank(P_train)-1

test_features= linear_subspace_projection(P_test, model, 1); P_train=model.train'; P_test=test_features';

%% Normalisation

P_train=P_train/256;
P_test=P_test/256;

%% % P_train=mapminmax(P_train,0,1); % P_test=mapminmax(P_test,0,1);

% %%%%%%%% load label %%%%%%%%%%%% train_label=load('train_label.txt'); test_label=load('test_label.txt');

% P_train = P_train'; % P_test = P_test';

%% classification K Nearest Neighour

% fit a kNN classification model %mdl = fitcknn(P_train,train_label,'Distance','euclidean','NumNeighbors',1); %euclidean

 Model = fitctree(P_train(:,1:20),train_label,'ClassNames',{'1','2','3','4','5','6','7','8','9','10','11','12','13','14','15','16','17','18','19','20'});
 [predictlabel,score] = resubPredict(Model);

%% Dorsal hand vein recognition using SVM

clc clear all; c=[]; addpath train; addpath test; mapping=getmapping(8,'u2'); %LBP

       W=[0,0,0,0,0,0,0; ...
          0,1,1,1,1,1,0; ...
          0,1,2,4,2,1,0; ...
          0,1,4,4,4,1,0; ...
          0,1,2,4,2,1,0; ...
          0,1,1,1,1,1,0; ...
          0,0,0,0,0,0,0];

 %%Training images

 for folder_idx = 1 : 20             % no of classes
 for i = 1 : 9                             % no of images per class

 thisfile = fullfile('train', num2str(folder_idx), [num2str(i) '.bmp ']); 
 B = imread(thisfile ); 
 X = double(B);
  X = imresize(X,[60 60],'bilinear');  
  H2=DSLBP(X,mapping,W);
  Gray=X;
  Gray=(Gray-mean(Gray(:)))/std(Gray(:))*20+60;
  lpqhist=lpq(Gray,3,1,1,'nh');       
  % imshow(lpqhist);
  a=[H2,lpqhist];
  c=[c;a];
  disp(sprintf('Done',i));

 end 
 end

%% Testing images

d=[];

for folder_idx = 1 : 20 %no of classes for i = 1 : 3 % no of images per class

 thisfile = fullfile('test', num2str(folder_idx), [num2str(i) '.bmp ']); 
 B = imread(thisfile ); 
  X = double(B);
  X = imresize(X,[60 60],'bilinear');  
  H2=DSLBP(X,mapping,W);
  Gray=X;
  Gray=(Gray-mean(Gray(:)))/std(Gray(:))*20+60;
  lpqhist=lpq(Gray,3,1,1,'nh');       
  a=[H2,lpqhist];
  d=[d;a];
  disp(sprintf('Done',i));
 end 
 end

%% P_train=c; P_test=d;

% %% PCA low dimension reduction % P_train = P_train';

% if classes are 20 then eiganvectors not exceed then 179 model = perform_pca(P_train,179); %rank(P_train)-1

test_features= linear_subspace_projection(P_test, model, 1); P_train=model.train'; P_test=test_features';

%% Normalisation

P_train=P_train/256;
P_test=P_test/256;

%% % P_train=mapminmax(P_train,0,1); % P_test=mapminmax(P_test,0,1);

% %%%%%%%% load label %%%%%%%%%%%% train_label=load('train_label.txt'); test_label=load('test_label.txt');

% P_train = P_train'; % P_test = P_test';

%% classification K Nearest Neighour

% fit a kNN classification model %mdl = fitcknn(P_train,train_label,'Distance','euclidean','NumNeighbors',1); %euclidean

 Model = fitctree(P_train(:,1:20),train_label,'ClassNames',{'1','2','3','4','5','6','7','8','9','10','11','12','13','14','15','16','17','18','19','20'});
 [predictlabel,score] = resubPredict(Model);

[Atheer Alkubeyyer]

I have used the fitcknn and perfcurve function to get the FPR and TPR. However, it only returns three values for both FPR and TPR, this is when using NumNeighbors =1, how to get the values of FPR and TPR for all instances?

[Walter Roberson]

Unfortunately, Balaji's code is poorly formated in a way that makes it ambiguous about whether some parts of it are commented out or not.

[naved alam]

[predictlabel,score,cost] = predict(mdl,P_test);