In order to use Sequential Minimal Optimization (SMO) for a SVM model, we can take the following steps:
- Split of data using "cvpartition". If you partition your data into training and testing set using "cvpartition", you are enabling the data for cross validation. Find more about this on the following documentation page: https://www.mathworks.com/help/stats/cvpartition.html. The following code can be used as a boilerplate to make the partitions:
% X: data, Y: labels
cv = cvpartition(size(X, 1), 'HoldOut', 0.3);
XTrain = X(training(cv), :);
YTrain = Y(training(cv), :);
XTest = X(test(cv), :);
YTest = Y(test(cv), :);
- Make an SVM using "fitcsvm". In this, you can keep the "solver" to be "SMO". More can be found on the documentation page: https://www.mathworks.com/help/stats/fitcsvm.html#bt9w6j6_sep_shared-Solver. Code to make the model and test it is as follows:
SVMModel = fitcsvm(XTrain, YTrain, 'KernelFunction', 'linear', 'Standardize', true, 'Solver', 'SMO');
YPred = predict(SVMModel, XTest);
% Calculate the accuracy
accuracy = sum(YPred == YTest) / length(YTest);
fprintf('Test Accuracy: %.2f%%\n', accuracy * 100);
Feel free to ask any follow-ups in case you need any more help.
