Contents

Description of demo_binaryclass_alt.m

Uses regression or multiclass submodels to do binary classification task

clear all
close all
generateData_curved

usage of regression binary classification

options_rg = [];
options_rg.subModel = @ml_regression_Huber;
options_rg.subOptions = [];
options_rg.epsilon = 0.3;
model_rg = ml_binaryclass_regression(Xtrain, ytrain, options_rg);
yhat_rg = model_rg.predict(model_rg, Xtest);
testError_rg = mean(yhat_rg ~= ytest);
fprintf('Averaged misclassification test error with %s is: %.3f\n', ...
        model_rg.name, testError_rg);
figure;
plot2DClassifier(Xtrain, ytrain, model_rg);

Averaged misclassification test error with Binary Classification by Huber Loss Linear Regression \epsilon=0.9 is: 0.080

usage of classification binary classification

options_cl = [];
options_cl.subModel = @ml_multiclass_GDA;
model_cl = ml_binaryclass_multiclass(Xtrain, ytrain, options_cl);
yhat_cl = model_cl.predict(model_cl, Xtest);
testError_cl = mean(yhat_cl ~= ytest);
fprintf('Averaged misclassification test error with %s is: %.3f\n', ...
        model_cl.name, testError_cl);
figure;
plot2DClassifier(Xtrain, ytrain, model_cl);

Averaged misclassification test error with Bin. Class., Discr. Classification: Generative Gaussian Model is: 0.049

Published with MATLAB® R2015a