一、环境准备
1. 工具箱安装
% 下载并安装libsvm-mat工具箱（推荐使用林教授版本）
% 解压后添加到MATLAB路径
addpath(genpath('libsvm-mat-2.91'));% 验证安装
version -libsvm
2. 数据准备
% 加载示例数据（鸢尾花数据集）
load fisheriris
X = meas(:,1:2); % 使用前两个特征
Y = grp2idx(species); % 类别标签% 数据标准化
[X_scaled, mu, sigma] = zscore(X);
二、核心SVM实现
1. 模型训练
% 基本训练代码
model = svmtrain(Y, X_scaled, '-t 2 -c 1 -g 0.1');% 保存模型
save('svm_model.mat', 'model');
2. 模型预测
% 加载测试数据
load('test_data.mat');
X_test_scaled = zscore(X_test);% 预测
[predict_label, accuracy, dec_values] = svmpredict(Y_test, X_test_scaled, model);
三、GUI开发实现
1. 界面设计（使用GUIDE）
% 创建GUI组件
fig = uifigure('Name','SVM GUI','Position',[100,100,600,400]);
btn_load = uibutton(fig,'Text','加载数据','Position',[20,300,100,30],'ButtonPushedFcn',@(btn,event) load_data());% 数据展示区域
ax = uiaxes(fig,'Position',[0.2,0.2,0.6,0.6]);
xlabel(ax,'特征1'); ylabel(ax,'特征2');% 参数设置面板
panel_params = uipanel(fig,'Title','参数设置','Position',[0.75,0.3,0.2,0.5]);
edit_c = uieditfield(panel_params,'numeric','Position',[10,20,80,25],'Label','C值:');
edit_gamma = uieditfield(panel_params,'numeric','Position',[10,50,80,25],'Label','Gamma:');
2. 回调函数实现
function load_data()% 数据加载回调[filename, pathname] = uigetfile({'*.mat','MAT文件';'*.csv','CSV文件'});if isequal(filename,0)return;enddata = load(fullfile(pathname,filename));global X Y;X = data(:,1:end-1);Y = data(:,end);% 数据可视化gscatter(X(:,1), X(:,2), Y);title('原始数据分布');
endfunction train_model()% 训练回调c = str2double(edit_c.Value);gamma = str2double(edit_gamma.Value);cmd = sprintf('-t 2 -c %f -g %f', c, gamma);model = svmtrain(Y, X, cmd);% 显示结果msgbox(sprintf('训练完成！准确率: %.2f%%', model.acc(1)));
end
四、关键功能扩展
1. 参数网格搜索
function grid_search()% 参数范围设置C_values = [0.1, 1, 10];gamma_values = [0.01, 0.1, 1];best_acc = 0;best_params = struct();for c = C_valuesfor g = gamma_valuescmd = sprintf('-t 2 -c %f -g %f', c, g);[~, ~, ~, acc] = svmpredict(Y, X, model, cmd);if acc(1) > best_accbest_acc = acc(1);best_params.C = c;best_params.Gamma = g;endendend% 显示最优参数msgbox(sprintf('最优参数: C=%.2f, Gamma=%.2f\n准确率=%.2f%%',...best_params.C, best_params.Gamma, best_acc));
end
2. 可视化模块
function plot_decision_boundary()% 绘制决策边界d = 0.02;[x1Grid, x2Grid] = meshgrid(min(X(:,1)):d:max(X(:,1)), ...min(X(:,2)):d:max(X(:,2)));grid = [x1Grid(:), x2Grid(:)];[~, scores] = svmpredict(zeros(size(grid,1),1), grid, model);[~, ~, ~, dec_values] = svmpredict(zeros(size(grid,1),1), grid, model);figure;gscatter(X(:,1), X(:,2), Y);hold on;contour(x1Grid, x2Grid, reshape(dec_values(:,2), size(x1Grid)), [0 0], 'k');title('SVM决策边界');legend('Location','best');hold off;
end
五、工程化优化
1. 大数据集处理
% 分块训练（适用于>10万样本）
batch_size = 1000;
n_batches = ceil(size(X,1)/batch_size);model = [];
for i = 1:n_batchesstart_idx = (i-1)*batch_size +1;end_idx = min(i*batch_size, size(X,1));X_batch = X(start_idx:end_idx,:);Y_batch = Y(start_idx:end_idx);model = svmtrain(Y_batch, X_batch, cmd, model);
end
2. GPU加速
% 使用gpuArray加速计算
if canUseGPUX_gpu = gpuArray(X_scaled);model = svmtrain(Y, X_gpu, cmd);model = gather(model);
end
参考代码 基于libsvm的支持向量机在MATLAB文件及其在MATLAB上的GUI www.youwenfan.com/contentcnk/65365.html
六、典型应用案例
1. 图像分类（手写数字识别）
% 加载MNIST数据集
[X, Y] = load_mnist();% 特征降维
[coeff, score, ~] = pca(X);
X_pca = score(:,1:20);% 训练SVM模型
model = svmtrain(Y, X_pca, '-t 0 -c 10');
2. 生物信息学（基因表达数据分析）
% 加载基因数据
load('gene_expression.mat');% 处理不平衡数据
pos_idx = find(Y==1); neg_idx = find(Y==0);
X_balanced = [X(pos_idx,:); X(neg_idx(1:1000),:)];
Y_balanced = [Y(pos_idx); Y(neg_idx(1:1000))];% 加权SVM训练
model = svmtrain(Y_balanced, X_balanced, '-w1 10 -t 2');
)
