Doing something simply dumb :(( Trying to get a meshgrid to read my third Class category but not working.

So Im using machine learning to read my training and testing data where X1 is my first feature, X2 is my second feature, and y is the category it belongs to. This is really simple but im having trouble with it.

Im saying in my code that one category is green (this is where y = 1), another category is blue (where y=2), and the last category is red (y=3). I can plot green and blue but not red data points over my mesh grid and i think this line of code is the problem:

    %%%%%%%   this is an examply of my y:   %%%%% 
    % y = [ones(60,1); 2*ones(60,1); 3*ones(60,1)];
    Y = data.y;
    Y = ismember(training_data.y,Y);
    scatter(training_data.X1(Y==1),training_data.X2(Y==1),'o' , 'MarkerEdgeColor', 'black', 'MarkerFaceColor', 'green');
    Y = ismember(training_data.y,Y);
    scatter(training_data.X1(Y==2),training_data.X2(Y==2) , 'o' , 'MarkerEdgeColor', 'black', 'MarkerFaceColor', 'red');
    Y = ismember(training_data.y,Y);
    scatter(training_data.X1(Y==3),training_data.X2(Y==3) , 'o' , 'MarkerEdgeColor', 'black', 'MarkerFaceColor', 'red');
    xlabel('X1');

This is the code for only 2 categories, and this runs with no issues.

labels = data.y;
Y = ismember(training_data.y,labels(1));
scatter(training_data.X1(Y),training_data.X2(Y), 'o' , 'MarkerEdgeColor', 'black', 'MarkerFaceColor', 'green');
scatter(training_data.X1(~Y),training_data.X2(~Y) , 'o' , 'MarkerEdgeColor', 'black', 'MarkerFaceColor', 'red');

Here is the full code:

%Read in the data and structure a table and rename the variables for
%simplicity
data = table([x(:,1)], [x(:,2)], [y]);
data.Properties.VariableNames = {'X1','X2','y'};%Read in the data and structure a table and rename the variables for


classifier_name = {'Naive Bayes','Decision Tree','Discriminant Analysis','Nearest Neighbor'};

classifier{1} = fitcnb(data,'y~X1+X2');

%SVM has trouble with more than 1 class
% classifier{2} = fitcsvm(x,y,'KernelFunction','polynomial','PolynomialOrder',6);

classifier{2} = fitctree(data,'y~X1+X2');

classifier{3} = fitcdiscr(data,'y~X1+X2');

% classifier{4} = fitcknn(x,y);
k = 5;
classifier{4} = fitcknn(data,'y~X1+X2','NumNeighbors',k,'Distance','euclidean');

% %%%%%%%%%%%%%%%%%%%%%% Loop through each Machine learning classifier,
% then specify accuracy
 for i = 1:numel(classifier)
   cv = cvpartition(classifier{i}.NumObservations,'HoldOut',0.2); %split into 20% testing and 80% training
   cross_validated_model = crossval(classifier{i},'cvpartition',cv);


%Create Predictions for training and test sets 
Predictions_test = predict(cross_validated_model.Trained{1},data(test(cv),1:end-1));
Predictions_train = predict(cross_validated_model.Trained{1},data(training(cv),1:end-1));

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%% Create Predictions %%%%
Predict = predict(cross_validated_model.Trained{1},data(test(cv),1:end-1));

%%%% Draw Confusion Matrix %%%% this sets up an accuracy based matrix
Result = confusionmat(cross_validated_model.Y(test(cv)),Predict);

%%% Accuracy %%%% calulates the accuracy based on the confusion matrix
display(classifier_name{i})
Testing_accuracy = ((Result(2,2)+Result(1,1))/((Result(2,2)+Result(1,1)+Result(1,2)+Result(2,1))))

%%%% Create Predictions %%%%
Predict = predict(cross_validated_model.Trained{1},data(training(cv),1:end-1));

%%%% Draw Confusion Matrix %%%% this sets up an accuracy based matrix
Result = confusionmat(cross_validated_model.Y(training(cv)),Predict);

%%% Accuracy %%%% calulates the accuracy based on the confusion matrix
display(classifier_name{i})
Training_accuracy = ((Result(2,2)+Result(1,1))/((Result(2,2)+Result(1,1)+Result(1,2)+Result(2,1))))

   
   
   labels = data.y;
    X1_range = min(data.X1(training(cv)))-1:1000:max(data.X1(training(cv)))+1;
    X2_range = min(data.X2(training(cv)))-1:1000:max(data.X2(training(cv)))+1;
    [xx1, xx2] = meshgrid(X1_range,X2_range);
    XGrid = [xx1(:) xx2(:)];
    predictions_meshgrid = predict(cross_validated_model.Trained{1},XGrid);
    %figure
    subplot(2,2,i);
    gscatter(xx1(:), xx2(:), predictions_meshgrid,'gbr');
    hold on
    training_data = data(training(cv),:);
    Y = data.y;
    Y = ismember(training_data.y,labels(1));
    scatter(training_data.X1(Y),training_data.X2(Y),'o' , 'MarkerEdgeColor', 'black', 'MarkerFaceColor', 'green');
    Y = ismember(training_data.y,labels(2));
    scatter(training_data.X1(~Y),training_data.X2(~Y) , 'o' , 'MarkerEdgeColor', 'black', 'MarkerFaceColor', 'blue');
    Y = ismember(training_data.y,labels(3));
    scatter(training_data.X1(Y==3),training_data.X2(Y==3) , 'o' , 'MarkerEdgeColor', 'black', 'MarkerFaceColor', 'red');
    xlabel('X1');
    ylabel('X2');
    title(classifier_name);
    legend off, axis tight
    % legend(labels,'Location',[0.45,0.01,0.45,0.05],'Orientation','Horizontal');
    hold off
   
%    subplot(2,2,i);
%    gscatter(xx1(:), xx2(:), predictedhealth,'gbr','osd');

   title(classifier_name{i})
   legend off, axis tight
end

Pretty Please help!

Answer 1

[yanqi liu]

yes，sir，may be use

index1 = find(Y==1);
index2 = find(Y==2);
index3 = find(Y==3);

to generate data index vector，and get data split to plot