[求助] matlab libsvm中svm参数优化时准确率超低，只有7%！求解决办法！！！

load data
>> load data_labels
>> train_data=[data(1:57,:)];
>> train_data_labels=[data_labels(1:57)];
>> test_data=[data(58:71,:)];
>> test_data_labels=[data_labels(58:71)];
>> [train_data,pstrain]=mapminmax(train_data');
>> pstrain.ymin=0;
>> patrain.ymax=1;
>> [train_data,pstrain]=mapminmax(train_data,pstrain);
>> test_data=[data(58:71,:)];
>> [test_data,pstest]=mapminmax(test_data');
>> pstest.ymin=0;
>> pstest.ymax=1;
>> [test_data,pstest]=mapminmax(test_data,pstest);
>> train_data=train_data';
>> test_data=test_data';
>> [bestacc,bestc,bestg] = SVMcgForClass(train_data_labels,train_data,-10,10,-10,10);
部分结果显示为：
optimization finished, #iter = 1
nu = 0.000977obj = -1.000000, rho = 0.000000nSV = 2, nBSV = 0
*optimization finished, #iter = 1
nu = 0.000977obj = -1.000000, rho = 0.000000nSV = 2, nBSV = 0
*optimization finished, #iter = 1
nu = 0.000977obj = -1.000000, rho = 0.000000nSV = 2, nBSV = 0
*optimization finished, #iter = 1
nu = 0.000977obj = -1.000000, rho = 0.000000nSV = 2, nBSV = 0Total nSV = 22
Cross Validation Accuracy = 7.01754%
求大神解答！！！！救命！！！一定要看到我！！！！！
下面是SVMcgForClass程序文件
function [ bestacc,bestc,bestg ] = SVMcgForClass(train_data_labels,train_data,cmin,cmax,gmin,gmax,v,cstep,gstep,accstep)
%UNTITLED3 Summary of this function goes here
%   Detailed explanation goes here
% SVMcgForClass
% 输入:
% train_data_labels:训练集标签.要求与libsvm工具箱中要求一致.
% train:训练集.要求与libsvm工具箱中要求一致.
% cmin:惩罚参数c的变化范围的最小值(取以2为底的对数后),即 c_min = 2^(cmin).默认为 -5
% cmax:惩罚参数c的变化范围的最大值(取以2为底的对数后),即 c_max = 2^(cmax).默认为 5
% gmin:参数g的变化范围的最小值(取以2为底的对数后),即 g_min = 2^(gmin).默认为 -5
% gmax:参数g的变化范围的最小值(取以2为底的对数后),即 g_min = 2^(gmax).默认为 5
% v:cross validation的参数,即给测试集分为几部分进行cross validation.默认为 3
% cstep:参数c步进的大小.默认为 1
% gstep:参数g步进的大小.默认为 1
% accstep:最后显示准确率图时的步进大小.默认为 1.5
% 输出:
% bestacc:Cross Validation 过程中的最高分类准确率
% bestc:最佳的参数c
% bestg:最佳的参数g
% about the parameters of SVMcgForClass
if nargin < 10
    accstep = 1.5;
end
if nargin < 8
    accstep = 1.5;
    cstep = 1;
    gstep = 1;
end
if nargin < 7
    accstep = 1.5;
    v = 3;
    cstep = 1;
    gstep = 1;
end
if nargin < 6
    accstep = 1.5;
    v = 3;
    cstep = 1;
    gstep = 1;
    gmax = 5;
end
if nargin < 5
    accstep = 1.5;
    v = 3;
    cstep = 1;
    gstep = 1;
    gmax = 5;
    gmin = -5;
end
if nargin < 4
    accstep = 1.5;
    v = 3;
    cstep = 1;
    gstep = 1;
    gmax = 5;
    gmin = -5;
    cmax = 5;
end
if nargin < 3
    accstep = 1.5;
    v = 3;
    cstep = 1;
    gstep = 1;
    gmax = 5;
    gmin = -5;
    cmax = 5;
    cmin = -5;
end
% X:c Y:g cg:accuracy
[X,Y] = meshgrid(cmin:cstep:cmax,gmin:gstep:gmax);
[m,n] = size(X);
cg = zeros(m,n);
% record accuracy with different c & g,and find the best accuracy with the smallest c
bestc = 0;
bestg = 0;
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_data_labels, train_data, cmd);
        if cg(i,j) > bestacc
            bestacc = cg(i,j);
            bestc = basenum^X(i,j);
            bestg = basenum^Y(i,j);
        end
        if ( cg(i,j) == bestacc && bestc > basenum^X(i,j) )
            bestacc = cg(i,j);
            bestc = basenum^X(i,j);
            bestg = basenum^Y(i,j);
        end
    end
end
% draw the accuracy with different c & g
figure;
[C,h] = contour(X,Y,cg,60:accstep:100);
clabel(C,h,'FontSize',20,'Color','r');
xlabel('log2c','FontSize',20);
ylabel('log2g','FontSize',20);
title('参数选择结果图(grid search)','FontSize',10);
grid on;

end

2楼: Originally posted by zgyzhp at 2017-07-07 21:26:34
上面的笑脸是：，不知道怎么变成这个了，更正一下，上面3行带笑脸的分别是
train_data=;  
test_data=;  
test_data=;
不要沉