libsvm使用实例

1 原始数据

samples.txt

+2 1:47817578 2:117348 3:41170351 4:201515

+2 1:39301334 2:156188 3:33958277 4:194650

+1 1:69723870 2:88085  3:62840950 4:209127

+1 1:71368734 2:88150  3:64478712 4:209395

+3 1:23745724 2:188299 3:18677611 4:185112

+2 1:40632238 2:119584 3:33491192 4:199092

+2 1:50372906 2:120723 3:43938211 4:201544

+2 1:38787418 2:128953 3:31769247 4:197855

+2 1:50824026 2:127297 3:43188486 4:200509

 

2 归一化数据(dos命令:svm-scale –l 0 –u 1 samples.txt > samples-scale.txt)

samples-scale.txt:

2 1:0.505467 2:0.292005 3:0.491096 4:0.675493

2 1:0.326641 2:0.679576 3:0.333631 4:0.392785

1 1:0.965461 3:0.964242 4:0.988963

1 1:1 2:0.000648612 3:1 4:1

3 2:1

2 1:0.354587 2:0.314317 3:0.323433 4:0.575711

2 1:0.559124 2:0.325683 3:0.551528 4:0.676687

2 1:0.315849 2:0.407807 3:0.285837 4:0.52477

2 1:0.568597 2:0.391283 3:0.535159 4:0.634065

 

3 训练样本(dos命令: svm-train samples-scale.txt train-result.txt)

  train-result.txt：

svm_type c_svc

kernel_type rbf

gamma 0.25

nr_class 3

total_sv 7

rho -0.701704 -0.864188 -0.375774

label 2 1 3

nr_sv 4 2 1

SV

0.453445662711471 0 1:0.505467 2:0.292005 3:0.491096 4:0.675493

0 1 1:0.326641 2:0.679576 3:0.333631 4:0.392785

1 0 1:0.559124 2:0.325683 3:0.551528 4:0.676687

0.5465543372885291 0 1:0.568597 2:0.391283 3:0.535159 4:0.634065

-1 1 1:0.965461 3:0.964242 4:0.988963

-1 0 1:1 2:0.000648612 3:1 4:1

-1 -1 2:1

 

4 预测

预测数据:temp.txt：

+2 1:47017578 2:107348 3:40170351 4:161515

预测dos命令:svm-predict temp.txt train-result.txt predict-result.txt

预测结果predict-reuslt.txt：

2

 

5结论：预测结果正确