libsvm（MATLAB版本）安装与使用

一.下载libsvm

1.libsvm - 3.22 : https://github.com/cjlin1/libsvm  配套使用MATLAB2013以上，VisualStudio2015以上

2.libsvm - 3.17 : http://download.csdn.net/download/wzh_xwjh/5648969  配套使用MATLAB2012a，VisualStudio2010

注：这里列的两个安装包是分别针对不同的MATLAB版本和编译器版本使用，否则会出现无法编译的错误。

二.安装libsvm

1.使用MATLAB版本的libsvm需要首先安装Visualstudio，安装过程中只需安装VisualC++即可，其他studio可以不用安装；

2.将libsvm解压后保存到MATLAB文件下的toolbox文件夹，然后打开MATLAB：

file - setpath - Add with Subfolders - 选中libsvm文件夹所在路径 - save - close

3. 打开matlab，设置当前路径为C:\Program Files\MATLAB\R2012a\toolbox\libsvm-3.17\matlab, 在command window中输入“mex -setup”，按提示输入y - 1/2/3/... - y

示例：

>> mex -setup
 
Welcome to mex -setup.  This utility will help you set up  
a default compiler.  For a list of supported compilers, see  
http://www.mathworks.com/support/compilers/R2012a/win64.html 
Please choose your compiler for building MEX-files: 
Would you like mex to locate installed compilers [y]/n? y
Select a compiler: 
[1] Microsoft Visual C++ 2010 in g:\VS\ 
[0] None 
Compiler: 1
Please verify your choices: 
Compiler: Microsoft Visual C++ 2010  
Location: g:\VS\ 
Are these correct [y]/n? y
*************************************************************************** 
  Warning: MEX-files generated using Microsoft Visual C++ 2010 require 
           that Microsoft Visual Studio 2010 run-time libraries be  
           available on the computer they are run on. 
           If you plan to redistribute your MEX-files to other MATLAB 
           users, be sure that they have the run-time libraries. 
*************************************************************************** 
Trying to update options file: C:\Users\liuyuchen\AppData\Roaming\MathWorks\MATLAB\R2012a\mexopts.bat 
From template:              C:\PROGRA~1\MATLAB\R2012a\bin\win64\mexopts\msvc100opts.bat 
 
Done . . . 
************************************************************************** 

4.在command window中输入“make”，完成编译

三.运行svmtrain与svmpredict

% 载入软件包数据
load heart_scale.mat;  %如果不加 .mat 程序默认数据为ASCII格式，出错：Error using load
Number of columns on line 3 of ASCII file C:\Program Files\MATLAB\R2012a\toolbox\libsvm-3.22\heart_scale
must be the same as previous lines. 
data = heart_scale_inst;
label = heart_scale_label;

% 选取前200个数据作为训练集合，后70个数据作为测试集合
ind = 200;
traindata = data(1:ind,:);
trainlabel = label(1:ind,:);
testdata = data(ind+1:end,:);
testlabel = label(ind+1:end,:);

% 利用训练集合建立分类模型
model = svmtrain(trainlabel,traindata,'-s 0 -t 2 -c 1.2 -g 2.8');

%参数输入的意义：
% -s svm类型：SVM设置类型(默认0)

% 0 -- C-SVC

% 1 --v-SVC
% 2 – 一类SVM
%3 -- e -SVR
% 4 -- v-SVR
% -t 核函数类型：核函数设置类型(默认2)
% 0 – 线性：u'v
% 1 – 多项式：(r*u'v + coef0)^degree
% 2 – RBF函数：exp(-r|u-v|^2)
% 3 –sigmoid：tanh(r*u'v + coef0)
% -g r(gama)：核函数中的gamma函数设置(针对多项式/rbf/sigmoid核函数)
% -c cost：设置C-SVC，e -SVR和v-SVR的参数(损失函数)(默认1)
% 分类模型model解密
model
Parameters = model.Parameters
Label = model.Label
nr_class = model.nr_class
totalSV = model.totalSV
nSV = model.nSV

% 利用建立的模型看其在训练集合上的分类效果
[ptrain,accuracy,prob_estimates] = svmpredict(trainlabel,traindata,model);

% 预测测试集合标签
[ptest,accuracy_test,prob_estimates_test] = svmpredict(testlabel,testdata,model);
 toc;

%%%%结果

.*
optimization finished, #iter = 350
nu = 0.719833
obj = -88.802168, rho = 0.058346
nSV = 197, nBSV = 13
Total nSV = 197
model = 
    Parameters: [5x1 double]
      nr_class: 2
       totalSV: 197
           rho: 0.0583
         Label: [2x1 double]
    sv_indices: [197x1 double]
         ProbA: []
         ProbB: []
           nSV: [2x1 double]
       sv_coef: [197x1 double]
           SVs: [197x13 double]
Parameters =
         0
    2.0000
    3.0000
    2.8000
         0
Label =
     1
    -1
nr_class =
     2
totalSV =
   197
nSV =
    89
   108
Accuracy = 99.5% (199/200) (classification)
Accuracy = 68.5714% (48/70) (classification)
Elapsed time is 0.013768 seconds.
optimization finished, #iter = 350
nu = 0.719833
obj = -88.802168, rho = 0.058346
nSV = 197, nBSV = 13
Total nSV = 197
model = 
    Parameters: [5x1 double]
      nr_class: 2
       totalSV: 197
           rho: 0.0583
         Label: [2x1 double]
    sv_indices: [197x1 double]
         ProbA: []
         ProbB: []
           nSV: [2x1 double]
       sv_coef: [197x1 double]
           SVs: [197x13 double]
Parameters =
         0
    2.0000
    3.0000
    2.8000
         0
Label =
     1
    -1
nr_class =
     2
totalSV =
   197
nSV =
    89
   108
Accuracy = 99.5% (199/200) (classification)
Accuracy = 68.5714% (48/70) (classification)
Elapsed time is 0.013768 seconds.