Felomeng翻译：libsvm2.88之实际使用中的技巧

<p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="font-size: small;"><span style="">数值化你的数据。例如，将所有的属性用区间</span><span style=""><span style="font-family: Times New Roman;">[0,1]</span></span><span style="">或</span><span style=""><span style="font-family: Times New Roman;">[-1,+1]</span></span><span style="">中的数值表示。</span></span></p>
<p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="font-size: small;"><span style=""><span style="font-family: Times New Roman;">*</span></span><span style="">使用</span><span style=""><span style="font-family: Times New Roman;">C-SVC</span></span><span style="">模型时，可以考虑选用“</span><span style=""><span style="font-family: Times New Roman;">tools</span></span><span style="">”文件夹中的模型选择工具。</span></span></p>
<p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="font-size: small;"><span style=""><span style="font-family: Times New Roman;">*</span></span><span style="">在</span><span style=""><span style="font-family: Times New Roman;">nu-SVC/one-class-SVM/nu-SVR</span></span><span style="">模型中的</span><span style=""><span style="font-family: Times New Roman;">nu</span></span><span style="">参数接近训练错误和支持向量的商。</span></span></p>
<p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="font-size: small;"><span style=""><span style="font-family: Times New Roman;">*</span></span><span style="">如果分类的目标数据是不均衡的（比如正数很多，且很少有负数），试着通过</span><span style=""><span style="font-family: Times New Roman;">-wi</span></span><span style="">参数调整惩罚因子</span><span style=""><span style="font-family: Times New Roman;">C</span></span><span style="">（参见下面的示例）。</span></span></p>
<p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="font-size: small;"><span style=""><span style="font-family: Times New Roman;">*</span></span><span style="">处理较大规模问题时设定较大的缓存大小（也就是扩大参数</span><span style=""><span style="font-family: Times New Roman;">-m</span></span><span style="">的数值）。</span></span></p>