%Create a decision tree, test it, and calculate accuracy/error

function [test_labels, train_labels, error, tree] = k_make_tree_script(train_patterns, train_targets, test_patterns, test_targets, depth)

ntest = length(test_targets);
ntrain = length(train_targets);

%Build tree
tree = k_make_tree( train_patterns, train_targets, depth); 

tree;

%Find labels (test)
test_labels = k_classify_w_tree( test_patterns, tree );

%Find labels (train)
train_labels = k_classify_w_tree( train_patterns, tree );


%Calculate accuracy, error
check_test = test_labels.' - test_targets;
num_incorrect_test = length(find(check_test ~= 0));
check_train = train_labels.' - train_targets;
num_incorrect_train = length(find(check_train ~= 0));
error(1,1) = num_incorrect_train / ntrain;
error(1,2) = num_incorrect_test / ntest;
end
% 
% plot(depths,error(:,1), '-ro', 'MarkerSize', 10 );  hold on; 
% plot(depths,error(:,2), '-bx', 'MarkerSize', 10 ); 
% grid on; 
% xlabel( 'tree depth' ); ylabel( 'error (%)' ); 

% %Plot decision boundaries
% 
% x1 = -2:.01:2; x2 = -2:.01:2;
% x1 = linspace(-2,+2,1500); x2 = linspace(-2,+2,1500);
% 
% x1 = linspace(-5,+5,1500); x2 = linspace(-5,+5,1500); 
% [X1,X2] = meshgrid(x1,x2);
% test_grid_pts = [X1(:) X2(:)]; 
% grid_labels = k_classify_w_tree( test_grid_pts, tree );
% figure; imagesc(x1,x2,reshape(grid_labels,length(x1),length(x2))); colorbar; 
% hold on; axis xy;
% 
% % (or)
% b = find(train_targets==0); 
% plot( train_patterns(b,1), train_patterns(b,2), 'wx' ); 
% g = find(train_targets==1); 
% plot( train_patterns(g,1), train_patterns(g,2), 'wo' ); 
% 
% % (or)
% b = find(test_targets==0); 
% plot( test_patterns(b,1), test_patterns(b,2), 'wx' );
% g = find(test_targets==1); 
% plot( test_patterns(g,1), test_patterns(g,2), 'wo' );