libSVM应用举例

 

      前面提到，很多人看到libSVM这么多的参数，估计要犯晕了。没关系，我之前把相关的libSVM参数已经讲解了一遍，这里，再给出libSVM的用法。如果你不想花时间去仔细研究libSVM，完全可以参照我的函数来直接调用libSVM完成你的工作。

      首先是训练SVM得到模型；假设，有10个训练样本，每个训练样本，有12个特征值，即：每个训练样本的维数是12，也就是说，训练样本构成了一个10*12的矩阵（当然，所有数据应该被归一化到[-1,1]或[0,1]范围内），另外，所有训练样本，应该有一个明确的类别标签，以此来表明当前样本所属的类别。所以，完整的训练样本，应该是10*13的矩阵，这里，我们给出一个10*13的矩阵，并认为，它就是训练样本。每个训练样本是个行向量，而该矩阵的第一列则代表训练样本的“标签”。即：第一个训练样本，类别标签为“1”，第二个训练样本类别标签为“-1”。第一个训练样本的第一个特征值为0.708333，第二个特征值为1，第三个特征值为1...

double inputArr[10][13] = 

{

1,0.708333,1,1,-0.320755,-0.105023,-1,1,-0.419847,-1,-0.225806,0,1, 

-1,0.583333,-1,0.333333,-0.603774,1,-1,1,0.358779,-1,-0.483871,0,-1,

1,0.166667,1,-0.333333,-0.433962,-0.383562,-1,-1,0.0687023,-1,-0.903226,-1,-1,

-1,0.458333,1,1,-0.358491,-0.374429,-1,-1,-0.480916,1,-0.935484,0,-0.333333,

-1,0.875,-1,-0.333333,-0.509434,-0.347032,-1,1,-0.236641,1,-0.935484,-1,-0.333333,

-1,0.5,1,1,-0.509434,-0.767123,-1,-1,0.0534351,-1,-0.870968,-1,-1,

1,0.125,1,0.333333,-0.320755,-0.406393,1,1,0.0839695,1,-0.806452,0,-0.333333,

1,0.25,1,1,-0.698113,-0.484018,-1,1,0.0839695,1,-0.612903,0,-0.333333,

1,0.291667,1,1,-0.132075,-0.237443,-1,1,0.51145,-1,-0.612903,0,0.333333,

1,0.416667,-1,1,0.0566038,0.283105,-1,1,0.267176,-1,0.290323,0,1

};

另外，我们给出一个待分类的输入向量，并用行向量形式表示：

double testArr[]=

{

0.25,1,1,-0.226415,-0.506849,-1,-1,0.374046,-1,-0.83871,0,-1

};

下面，给出完整的SVM训练及预测程序：

#include "stdafx.h"

#include "svm.h"

#include "iostream"

#include "fstream"

using namespace std;  

 

double inputArr[10][13] = 

{

1,0.708333,1,1,-0.320755,-0.105023,-1,1,-0.419847,-1,-0.225806,0,1, 

-1,0.583333,-1,0.333333,-0.603774,1,-1,1,0.358779,-1,-0.483871,0,-1,

1,0.166667,1,-0.333333,-0.433962,-0.383562,-1,-1,0.0687023,-1,-0.903226,-1,-1,

-1,0.458333,1,1,-0.358491,-0.374429,-1,-1,-0.480916,1,-0.935484,0,-0.333333,

-1,0.875,-1,-0.333333,-0.509434,-0.347032,-1,1,-0.236641,1,-0.935484,-1,-0.333333,

-1,0.5,1,1,-0.509434,-0.767123,-1,-1,0.0534351,-1,-0.870968,-1,-1,

1,0.125,1,0.333333,-0.320755,-0.406393,1,1,0.0839695,1,-0.806452,0,-0.333333,

1,0.25,1,1,-0.698113,-0.484018,-1,1,0.0839695,1,-0.612903,0,-0.333333,

1,0.291667,1,1,-0.132075,-0.237443,-1,1,0.51145,-1,-0.612903,0,0.333333,

1,0.416667,-1,1,0.0566038,0.283105,-1,1,0.267176,-1,0.290323,0,1

};

 

double testArr[]=

{

0.25,1,1,-0.226415,-0.506849,-1,-1,0.374046,-1,-0.83871,0,-1

};

 

void DefaultSvmParam(struct svm_parameter *param)

{

param->svm_type = C_SVC;

param->kernel_type = RBF;

param->degree = 3;

param->gamma = 0; // 1/num_features

param->coef0 = 0;

param->nu = 0.5;

param->cache_size = 100;

param->C = 1;

param->eps = 1e-3;

param->p = 0.1;

param->shrinking = 1;

param->probability = 0;

param->nr_weight = 0;

param->weight_label = NULL;

param->weight = NULL;

}

 

void SwitchForSvmParma(struct svm_parameter *param, char ch, char *strNum, int nr_fold, int cross_validation)

{

switch(ch)

{

case 's':

{

param->svm_type = atoi(strNum);

break;

}

case 't':

{

param->kernel_type = atoi(strNum);

break;

}

case 'd':

{

param->degree = atoi(strNum);

break;

}

case 'g':

{

param->gamma = atof(strNum);

break;

}

case 'r':

{

param->coef0 = atof(strNum);

break;

}

case 'n':

{

param->nu = atof(strNum);

break;

}

case 'm':

{

param->cache_size = atof(strNum);

break;

}

case 'c':

{

param->C = atof(strNum);

break;

}

case 'e':

{

param->eps = atof(strNum);

break;

}

case 'p':

{

param->p = atof(strNum);

break;

}

case 'h':

{

param->shrinking = atoi(strNum);

break;

}

case 'b':

{

param->probability = atoi(strNum);

break;

}

case 'q':

{

break;

}

case 'v':

{

cross_validation = 1;

nr_fold = atoi(strNum);

if (nr_fold < 2)

{

cout<<"nr_fold should > 2!!! file: "<<__FILE__<<" function: ";

cout<<__FUNCTION__<<" line: "<<__LINE__<<endl;

 

}

break;

}

case 'w':

{

++param->nr_weight;

param->weight_label = (int *)realloc(param->weight_label,sizeof(int)*param->nr_weight);

param->weight = (double *)realloc(param->weight,sizeof(double)*param->nr_weight);

 

param->weight_label[param->nr_weight-1] = atoi(strNum);

 

param->weight[param->nr_weight-1] = atof(strNum);

break;

}

default:

{

break;

}

}

}

 

void SetSvmParam(struct svm_parameter *param, char *str, int cross_validation, int nr_fold)

{

DefaultSvmParam(param);

cross_validation = 0;

 

char ch = ' ';

int strSize = strlen(str);

for (int i=0; i<strSize; i++)

{

if (str[i] == '-')

{

ch = str[i+1];

int length = 0;

for (int j=i+3; j<strSize; j++)

{

if (isdigit(str[j]))

{

length++;

}

else

{

break;

}

}

char *strNum = new char[length+1];

int index = 0;

for (int j=i+3; j<i+3+length; j++)

{

strNum[index] = str[j];

index++;

}

strNum[length] = '/0';

SwitchForSvmParma(param, ch, strNum, nr_fold, cross_validation);

delete strNum;

}

}

}

 

void SvmTraining(char *option)

{

struct svm_parameter param;

struct svm_problem prob;

struct svm_model *model;

struct svm_node *x_space;

int cross_validation = 0;

int nr_fold = 0;

 

int sampleCount = 10;

int featureDim = 12;

prob.l =  sampleCount;

prob.y = new double[sampleCount];

prob.x = new struct svm_node*[sampleCount];

x_space = new struct svm_node[(featureDim+1)*sampleCount];

 

SetSvmParam(&param, option, cross_validation, nr_fold);

for (int i=0; i<sampleCount; i++)

{

prob.y[i] = inputArr[i][0];

}

 

int j = 0;

for (int i=0; i<sampleCount; i++)

{

prob.x[i] = &x_space[j];

for (int k=1; k<=featureDim; k++)

{

x_space[i*featureDim+k].index = k;

x_space[i*featureDim+k].value = inputArr[i][k];

}

x_space[(i+1)*featureDim].index = -1;

j = (i+1)*featureDim + 1; 

}

 

 

model = svm_train(&prob, &param);

const char* model_file_name = "C://Model.txt";

svm_save_model(model_file_name, model);

svm_destroy_model(model);

svm_destroy_param(&param);

delete[] prob.y;

delete[] prob.x;

delete[] x_space; 

}

int SvmPredict(const char* modelAdd)

{

struct svm_node *testX;

struct svm_model* testModel;

testModel = svm_load_model(modelAdd);

int featureDim = 12;

testX = new struct svm_node[featureDim+1];

for (int i=0; i<featureDim; i++)

{

testX[i].index = i+1;

testX[i].value = testArr[i];

}

testX[featureDim].index = -1;

double p = svm_predict(testModel, testX);

svm_destroy_model(testModel);

delete[] testX;

if (p > 0.5)

{

return 1;

}

else

{

return -1;

}

}

int _tmain(int argc, _TCHAR* argv[])

{

SvmTraining("-c 100 -t 1 -g 4 -r 1 -d 4");

int flag = SvmPredict("c://model.txt");

cout<<"flag = "<<flag<<endl;

system("pause");

return 0;

}