confusion matrix

Author: avouros

unused/init_results_confusion_matrix.m

@@ -0,0 +1,11 @@
+ppath = 'I:\Documents\MWMGEN\tiago_original\Mclassification\class_1301_10388_250_07_10_10_mr0';
+output_dir = '\\staffstore\avgoustinos\.redirect\Desktop\res';
+
+folds = 10;
+files = dir(fullfile(ppath,'*.mat'));
+for i = 1:length(files)
+    output_dir = fullfile(output_dir,files(i).name);
+    mkdir(output_dir);
+    load(fullfile(ppath,files(i).name));
+    results_confusion_matrix(classification_configs,folds,output_dir);
+end

unused/results_confusion_matrix.asv

@@ -0,0 +1,76 @@
+function results_confusion_matrix(segmentation_configs,classification_configs,folds,output_dir)
+% Computes the confusion matrix for the classification of segments.
+% Values are the total number of missclassifications for a 10-fold
+% cross-validation of the clustering algorithm
+        
+    fn = fullfile(output_dir, 'confusion_matrix.mat');
+    fn2 = fullfile(output_dir, 'Confusion_matrix.txt');
+
+    % get classifier object
+    classif = classification_configs.SEMISUPERVISED_CLUSTERING;
+    defaults_clusters = classification_configs.DEFAULT_NUMBER_OF_CLUSTERS;
+
+    % check if the folds number is too high
+    a = length(classification_configs.CLASSIFICATION.class_map);
+    if iscell(folds)
+        if isempty(str2num(folds{1,1}))
+            disp('Insert a number to specify the N-fold cross-validation.');
+            return
+        end    
+    elseif ~isnumeric(folds)  
+        if isempty(str2num(folds))
+            disp('Insert a number to specify the N-fold cross-validation.');
+            return
+        end    
+        if a - folds <= 0 || a - folds < a/10
+            disp('Input number too high, insert a lower number.');
+            return
+        end
+    else
+        if a - folds <= 0 || a - folds < a/10
+            disp('Input number too high, insert a lower number.');
+            return
+        end
+    end    
+    
+    % perform a N-fold cross-validation
+    res = classif.cluster_cross_validation(defaults_clusters, 'Folds', folds);
+
+    % take the "total confusion matrix"    
+    cm = res.results(1).confusion_matrix;
+    for i = 2:folds
+        cm = cm + res.results(i).confusion_matrix;
+    end
+    tags = res.results(1).classes;
+    
+    % NaN values
+    for i = 1:size(cm,1)
+        for j = 1:size(cm,2)
+            if isnan(cm)
+                cm(i,j) = 0;
+            end
+        end
+    end    
+
+    % save data
+    save(fn, 'tags', 'cm');   
+    disp('Tags:');
+    for i = 1:length(tags)
+        fprintf('%s\n', tags{1,i}{1,2});
+    end  
+    fprintf('\nConfusion matrix:\n');    
+    cm        
+    
+    % save to file
+    fileID = fopen(fn2,'wt');
+    for i = 1:length(tags)
+        fprintf(fileID,'%d. %s\n', i, tags{1,i}{1,2});
+    end
+    fprintf(fileID, '\n');
+    for i=1:size(cm,1)
+        fprintf(fileID, '%d ', cm(i,:));
+        fprintf(fileID, '\n');
+    end
+    fclose(fileID);
+    
+end

unused/results_confusion_matrix.m

@@ -1,38 +1,15 @@
-function results_confusion_matrix(segmentation_configs,classification_configs,folds)
+function results_confusion_matrix(classification_configs,folds,output_dir)
 % Computes the confusion matrix for the classification of segments.
-% Values are the total number of missclassifications for a 10-fold
+% Values are the total number of missclassifications for a N-fold
 % cross-validation of the clustering algorithm
 
-    fn = fullfile(strcat(segmentation_configs.OUTPUT_DIR,'/'), 'confusion_matrix.mat');
-    fn2 = fullfile(strcat(segmentation_configs.OUTPUT_DIR,'/'), 'Confusion_matrix.txt');
+    fn = fullfile(output_dir, 'confusion_matrix.mat');
+    fn2 = fullfile(output_dir, 'Confusion_matrix.txt');
 
     % get classifier object
     classif = classification_configs.SEMISUPERVISED_CLUSTERING;
     defaults_clusters = classification_configs.DEFAULT_NUMBER_OF_CLUSTERS;
 
-    % check if the folds number is too high
-    a = length(classification_configs.CLASSIFICATION.class_map);
-    if iscell(folds)
-        if isempty(str2num(folds{1,1}))
-            disp('Insert a number to specify the N-fold cross-validation.');
-            return
-        end    
-    elseif ~isnumeric(folds)  
-        if isempty(str2num(folds))
-            disp('Insert a number to specify the N-fold cross-validation.');
-            return
-        end    
-        if a - folds <= 0 || a - folds < a/10
-            disp('Input number too high, insert a lower number.');
-            return
-        end
-    else
-        if a - folds <= 0 || a - folds < a/10
-            disp('Input number too high, insert a lower number.');
-            return
-        end
-    end    
-    
     % perform a N-fold cross-validation
     res = classif.cluster_cross_validation(defaults_clusters, 'Folds', folds);