libsvm支持向量机回归

libsvm支持向量机算法包的基本使用,此处演示的是支持向量回归机，代码如下：

import java.io.BufferedReader;
import java.io.File;
import java.io.FileReader;
import java.util.ArrayList;
import java.util.List;
import libsvm.svm;
import libsvm.svm_model;
import libsvm.svm_node;
import libsvm.svm_parameter;
import libsvm.svm_problem;
public class SVM {
 public static void main(String[] args) {
  // 定义训练集点a{10.0, 10.0} 和 点b{-10.0, -10.0}，对应lable为{1.0, -1.0}
  List<Double> label = new ArrayList<Double>();
  List<svm_node[]> nodeSet = new ArrayList<svm_node[]>();
  getData(nodeSet, label, "file/train.txt");


  int dataRange=nodeSet.get(0).length;
  svm_node[][] datas = new svm_node[nodeSet.size()][dataRange]; // 训练集的向量表
  for (int i = 0; i < datas.length; i++) {
   for (int j = 0; j < dataRange; j++) {
    datas[i][j] = nodeSet.get(i)[j];
   }
  }
  double[] lables = new double[label.size()]; // a,b 对应的lable
  for (int i = 0; i < lables.length; i++) {
   lables[i] = label.get(i);
  }
  // 定义svm_problem对象
  svm_problem problem = new svm_problem();
  problem.l = nodeSet.size(); // 向量个数
  problem.x = datas; // 训练集向量表
  problem.y = lables; // 对应的lable数组
  // 定义svm_parameter对象
  svm_parameter param = new svm_parameter();
  param.svm_type = svm_parameter.EPSILON_SVR;
  param.kernel_type = svm_parameter.LINEAR;
  param.cache_size = 100;
  param.eps = 0.00001;
  param.C = 1.9;
  // 训练SVM分类模型
  System.out.println(svm.svm_check_parameter(problem, param));
  // 如果参数没有问题，则svm.svm_check_parameter()函数返回null,否则返回error描述。
  svm_model model = svm.svm_train(problem, param);
  // svm.svm_train()训练出SVM分类模型
  // 获取测试数据
  List<Double> testlabel = new ArrayList<Double>();
  List<svm_node[]> testnodeSet = new ArrayList<svm_node[]>();
  getData(testnodeSet, testlabel, "file/test.txt");
  svm_node[][] testdatas = new svm_node[testnodeSet.size()][dataRange]; // 训练集的向量表
  for (int i = 0; i < testdatas.length; i++) {
   for (int j = 0; j < dataRange; j++) {
    testdatas[i][j] = testnodeSet.get(i)[j];
   }
  }
  double[] testlables = new double[testlabel.size()]; // a,b 对应的lable
  for (int i = 0; i < testlables.length; i++) {
   testlables[i] = testlabel.get(i);
  }
  // 预测测试数据的lable
  double err = 0.0;
  for (int i = 0; i < testdatas.length; i++) {
   double truevalue = testlables[i];
   System.out.print(truevalue + " ");
   double predictValue = svm.svm_predict(model, testdatas[i]);
   System.out.println(predictValue);
   err += Math.abs(predictValue - truevalue);
  }
  System.out.println("err=" + err / datas.length);
 }
 public static void getData(List<svm_node[]> nodeSet, List<Double> label,
   String filename) {
  try {
   FileReader fr = new FileReader(new File(filename));
   BufferedReader br = new BufferedReader(fr);
   String line = null;
   while ((line = br.readLine()) != null) {
    String[] datas = line.split(",");
    svm_node[] vector = new svm_node[datas.length - 1];
    for (int i = 0; i < datas.length - 1; i++) {
     svm_node node = new svm_node();
     node.index = i + 1;
     node.value = Double.parseDouble(datas[i]);
     vector[i] = node;
    }
    nodeSet.add(vector);
    double lablevalue = Double.parseDouble(datas[datas.length - 1]);
    label.add(lablevalue);
   }
  } catch (Exception e) {
   e.printStackTrace();
  }
 }
}

训练数据，最后一列为目标值

复制代码 代码如下:

17.6,17.7,17.7,17.7,17.8
17.7,17.7,17.7,17.8,17.8
17.7,17.7,17.8,17.8,17.9
17.7,17.8,17.8,17.9,18
17.8,17.8,17.9,18,18.1
17.8,17.9,18,18.1,18.2
17.9,18,18.1,18.2,18.4
18,18.1,18.2,18.4,18.6
18.1,18.2,18.4,18.6,18.7
18.2,18.4,18.6,18.7,18.9
18.4,18.6,18.7,18.9,19.1
18.6,18.7,18.9,19.1,19.3

测试数据

18.7,18.9,19.1,19.3,19.6
18.9,19.1,19.3,19.6,19.9
19.1,19.3,19.6,19.9,20.2
19.3,19.6,19.9,20.2,20.6
19.6,19.9,20.2,20.6,21
19.9,20.2,20.6,21,21.5
20.2,20.6,21,21.5,22

运行结果如下：

null
.....................*
optimization finished, #iter = 521
nu = 0.8563100606562879
obj = -0.8743284942789487, rho = 3.4446474688721462
nSV = 12, nBSV = 9
19.6 19.55027196559459
19.9 19.845547309399905
20.2 20.175593570159805
20.6 20.540410747880657
21.0 20.95576975951272
21.5 21.40589968810346
22.0 21.945908541795603
err=0.03171653479610518