SVM的参数优化——交叉验证选择最佳参数c&g

支持向量机（Support Vector Machine，SVM）是一种广泛应用的监督式机器学习算法。它主要用于分类任务，但也适用于回归任务。在本文中，我们将深入探讨支持向量机的两个重要参数：C和gamma。

数据：UCI Machine Learning Repository

 %% 子函数 SVMcgForClass.m

function [bestacc,bestc,bestg] = SVMcgForClass(train_label,train,cmin,cmax,gmin,gmax,v,cstep,gstep,accstep)
%SVMcg cross validation by faruto

%
% by faruto
%Email:patrick.lee@foxmail.com QQ:516667408 http://blog.sina.com.cn/faruto BNU
%last modified 2010.01.17
%Super Moderator @ www.ilovematlab.cn

% 若转载请注明：
% faruto and liyang , LIBSVM-farutoUltimateVersion 
% a toolbox with implements for support vector machines based on libsvm, 2009. 
% Software available at http://www.ilovematlab.cn
% 
% Chih-Chung Chang and Chih-Jen Lin, LIBSVM : a library for
% support vector machines, 2001. Software available at
% http://www.csie.ntu.edu.tw/~cjlin/libsvm

% about the parameters of SVMcg 
if nargin < 10
    accstep = 4.5;
end
if nargin < 8
    cstep = 0.8;
    gstep = 0.8;
end
if nargin < 7
    v = 5;
end
if nargin < 5
    gmax = 8;
    gmin = -8;
end
if nargin < 3
    cmax = 8;
    cmin = -8;
end
% X:c Y:g cg:CVaccuracy
[X,Y] = meshgrid(cmin:cstep:cmax,gmin:gstep:gmax);
[m,n] = size(X);
cg = zeros(m,n);

eps = 10^(-4);

% record acc with different c & g,and find the bestacc with the smallest c
bestc = 1;
bestg = 0.1;
bestacc = 0;
basenum = 2;
for i = 1:m
    for j = 1:n
        cmd = ['-v ',num2str(v),' -c ',num2str( basenum^X(i,j) ),' -g ',num2str( basenum^Y(i,j) )];
        cg(i,j) = svmtrain(train_label, train, cmd);
        
        if cg(i,j) <= 55
            continue;
        end
        
        if cg(i,j) > bestacc
            bestacc = cg(i,j);
            bestc = basenum^X(i,j);
            bestg = basenum^Y(i,j);
        end        
        
        if abs( cg(i,j)-bestacc )<=eps && bestc > basenum^X(i,j) 
            bestacc = cg(i,j);
            bestc = basenum^X(i,j);
            bestg = basenum^Y(i,j);
        end        
        
    end
end
% to draw the acc with different c & g
figure;
[C,h] = contour(X,Y,cg,70:accstep:100);
clabel(C,h,'Color','r');
xlabel('log2c','FontSize',12);
ylabel('log2g','FontSize',12);
firstline = 'SVC参数选择结果图(等高线图)[GridSearchMethod]'; 
secondline = ['Best c=',num2str(bestc),' g=',num2str(bestg), ...
    ' CVAccuracy=',num2str(bestacc),'%'];
title({firstline;secondline},'Fontsize',12);
grid on; 

figure;
meshc(X,Y,cg);
% mesh(X,Y,cg);
% surf(X,Y,cg);
axis([cmin,cmax,gmin,gmax,30,100]);
xlabel('log2c','FontSize',12);
ylabel('log2g','FontSize',12);
zlabel('Accuracy(%)','FontSize',12);
firstline = 'SVC参数选择结果图(3D视图)[GridSearchMethod]'; 
secondline = ['Best c=',num2str(bestc),' g=',num2str(bestg), ...
    ' CVAccuracy=',num2str(bestacc),'%'];
title({firstline;secondline},'Fontsize',12);

% %SVM的参数优化——交叉验证选择最佳参数c&g提升分类器的性能 

%% 清空环境变量

close all;
clear;
clc;
format compact;

%% 数据提取

% 载入测试数据wine,其中包含的数据为classnumber = 3,wine:178*13的矩阵,wine_labes:178*1的列向量
load wine.mat;

% 画出测试数据的box可视化图

figure;
boxplot(wine,'orientation','horizontal','labels',categories);
title('wine数据的box可视化图','FontSize',12);
xlabel('属性值','FontSize',12);
grid on;

% 画出测试数据的分维可视化图

figure
subplot(3,5,1);
hold on
for run = 1:178
    plot(run,wine_labels(run),'*');
end
xlabel('样本','FontSize',10);
ylabel('类别标签','FontSize',10);
title('class','FontSize',10);
for run = 2:14
    subplot(3,5,run);
    hold on;
    str = ['attrib ',num2str(run-1)];
    for i = 1:178
        plot(i,wine(i,run-1),'*');
    end
    xlabel('样本','FontSize',10);
    ylabel('属性值','FontSize',10);
    title(str,'FontSize',10);
end

% 选定训练集和测试集

% 将第一类的1-30,第二类的60-95,第三类的131-153做为训练集
train_wine = [wine(1:30,:);wine(60:95,:);wine(131:153,:)];
% 相应的训练集的标签也要分离出来
train_wine_labels = [wine_labels(1:30);wine_labels(60:95);wine_labels(131:153)];
% 将第一类的31-59,第二类的96-130,第三类的154-178做为测试集
test_wine = [wine(31:59,:);wine(96:130,:);wine(154:178,:)];
% 相应的测试集的标签也要分离出来
test_wine_labels = [wine_labels(31:59);wine_labels(96:130);wine_labels(154:178)];

%% 数据预处理

% 数据预处理,将训练集和测试集归一化到[0,1]区间

[mtrain,ntrain] = size(train_wine);
[mtest,ntest] = size(test_wine);

dataset = [train_wine;test_wine];
% mapminmax为MATLAB自带的归一化函数
[dataset_scale,ps] = mapminmax(dataset',0,1);
dataset_scale = dataset_scale';

train_wine = dataset_scale(1:mtrain,:);
test_wine = dataset_scale( (mtrain+1):(mtrain+mtest),: );

%% 选择最佳的SVM参数c&g

% 首先进行粗略选择: c&g 的变化范围是 2^(-10),2^(-9),...,2^(10)

[bestacc,bestc,bestg] = SVMcgForClass(train_wine_labels,train_wine,-10,10,-10,10);

% 打印粗略选择结果

disp('打印粗略选择结果');
str = sprintf( 'Best Cross Validation Accuracy = %g%% Best c = %g Best g = %g',bestacc,bestc,bestg);
disp(str);

% 根据粗略选择的结果图再进行精细选择: c 的变化范围是 2^(-2),2^(-1.5),...,2^(4), g 的变化范围是 2^(-4),2^(-3.5),...,2^(4),

[bestacc,bestc,bestg] = SVMcgForClass(train_wine_labels,train_wine,-2,4,-4,4,3,0.5,0.5,0.9);

% 打印精细选择结果

disp('打印精细选择结果');
str = sprintf( 'Best Cross Validation Accuracy = %g%% Best c = %g Best g = %g',bestacc,bestc,bestg);
disp(str);

%% 利用最佳的参数进行SVM网络训练

cmd = ['-c ',num2str(bestc),' -g ',num2str(bestg)];
model = svmtrain(train_wine_labels,train_wine,cmd);

%% SVM网络预测

[predict_label,accuracy,predict_values] = svmpredict(test_wine_labels,test_wine,model);

% 打印测试集分类准确率

total = length(test_wine_labels);
right = sum(predict_label == test_wine_labels);
disp('打印测试集分类准确率');
str = sprintf( 'Accuracy = %g%% (%d/%d)',accuracy(1),right,total);
disp(str);

%% 结果分析

% 测试集的实际分类和预测分类图
% 通过图可以看出只有三个测试样本是被错分的

figure;
hold on;
plot(test_wine_labels,'o');
plot(predict_label,'r*');
xlabel('测试集样本','FontSize',12);
ylabel('类别标签','FontSize',12);
legend('实际测试集分类','预测测试集分类');
title('测试集的实际分类和预测分类图','FontSize',12);
grid on;
snapnow;

源自：《Matlab神经网络43个案例分析》。