用libsvm进行回归预测

时间:2021-12-20 04:36:36

最近因工作需要,学习了*大学林智仁(Lin Chih-Jen)教授等人开发的SVM算法开源算法包。

为了以后方便查阅,特把环境配置及参数设置等方面的信息记录下来。

用libsvm进行回归预测

林教授年轻时照片

SVM属于十大挖掘算法之一,主要用于分类和回归。本文主要介绍怎么使用LIBSVM的回归进行数值预测。

LIBSVM内置了多种编程语言的接口,本文选择Python。

1  LIBSVM官方网址

aaarticlea/png;base64,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" alt="" />

http://www.csie.ntu.edu.tw/~cjlin/libsvm/

可在这里下载LIBSVM的开源包,特别推荐初学者阅读文章A practical guide to SVM classification 和开源包自带的readme文件。可解决你的很多疑问。

一份简单易懂的LibSVM的学习资料。  http://pan.baidu.com/s/1bnfNmv9

2  安装环境

开源包版本 LIBSVM-3.20

操作系统  Win7 64bit

Python版本:python2.7.9                        下载链接 https://www.python.org/downloads/

gnuplot: Gnuplot Version 5.0 (Jan 2015)   下载链接 http://sourceforge.net/projects/gnuplot/files/

3 回归预测

A practical guide to SVM classification推荐的方法流程是这样的。

aaarticlea/png;base64,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" alt="" />

需要说明的是,回归预测需要gridsearch三个参数 gamma 、cost和epsilon;具体意义见下图红框。

aaarticlea/png;base64,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" alt="" />

开源包自带的grid.py 文件是针对分类用的,回归需要用gridregression.py文件。该文件需要另外下载。另外附带一份介绍LIBSVM使用的材料。 下载链接 http://pan.baidu.com/s/1bnfNmv9

下载完成后,把gridregression.py文件中的svm-train和gnuplot的安装路径修改为自己主机的安装路径。一定要认真,笔者在这里浪费好多时间。

3.1 数据格式整理

aaarticlea/png;base64,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" alt="" />

3.2 归一化

回归预测需要对训练集trainset进行归一化,并对测试集testset进行同样的归一化。

$  svm-scale -y -1 1 -s scale train.txt > trainScale.txt

$ svm-scale -r scale test.txt > testScale.txt

-y 参数表示要对label进行归一化。可在cmd输入svm-scale 回车, 查看各参数的意义。

3.3 gridsearch 寻找最优参数

python gridregression.py -log2c -10,10,1 -log2g -10,10,1 -log2p -10,10,1 -v 10 -s 3 -t 2 trainScale.txt > trainrs.txt

-s 3 表示进行回归 -t 2 表示使用径向基核函数

后查看trainrs.txt

aaarticlea/png;base64,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" alt="" />

红框中的三个数字对应最好的cost 、gamma和epsilon;

3.4 训练trainScale

svm-train -s 3 -t 2 -c 64 -g 0.03125 -p 0.125 trainScale.txt

会生成trainScale.txt.model 文件

3.5 预测testScale

svm-predict testScale trainScale.txt.model  predict-result.txt

查看predict-result.txt内容

aaarticlea/png;base64,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" alt="" />

3.6 对结果反归一化

比如x属于原来的测试集, 范围是[min, max], 而scale后的范围是[m, n]

那么x对应归一化的值y是什么?y = (x-min)/(max-min) *(n-m) + m

那如果已知y, x又是多少呢?

x = (y-m)/(n-m) * (max-min) + min

可按照x = (y-m)/(n-m) * (max-min) + min 对predict-result.txt内容进行反归一化,从而得到最终的预测值。

4 model参数说明

用libsvm进行回归预测

其中, #iter 为迭代次数;

nu 是选择的核函数类型的参数;

obj 为SVM文件转换为的二次规划求解得到的最小值;

rho 为判决函数的偏置项b;

nSV 为标准支持向量个数(0<a[i]<c);

nBSV 为边界上的支持向量个数(a[i]=c);

Total nSV为支持向量总个数(对于两类来说, 因为只有一个分类模型Total nSV = nSV

但是对于多类, 这个是各个分类模型的nSV之和).

在目录下, 还可以看到产生了一个train.model文件, 可以用记事本打开, 记录了训练后的结果.

  1. svm_type c_svc             //所选择的svm类型, 默认为c_svc
  2. kernel_type rbf             //训练采用的核函数类型, 此处为RBF核
  3. gamma 0.0769231          //RBF核的参数γ
  4. nr_class 2                 //类别数, 此处为两分类问题
  5. total_sv 132               //支持向量总个数
  6. rho 0.424462              //判决函数的偏置项b
  7. label 1 -1                 //原始文件中的类别标识
  8. nr_sv 64 68               //每个类的支持向量机的个数
  9. SV                     //以下为各个类的权系数及相应的支持向量
  10. 1 1:0.166667 2:1 3:-0.333333 … 10:-0.903226 11:-1 12:-1 13:1
  11. 0.5104832128985164 1:0.125 2:1 3:0.333333 … 10:-0.806452 12:-0.333333 13:0.5
  12. ………
  13. -1 1:-0.375 2:1 3:-0.333333…. 10:-1 11:-1 12:-1 13:1
  14. -1 1:0.166667 2:1 3:1 …. 10:-0.870968 12:-1 13:0.5

这里注意, 第二行出现的权系数为小数(0.5104832128985164)是因为这个点属于非边界上的支持向量, 即: (0<a[i]<c).

其他的两个(svm-predict, svm-scale)的使用过程类似.

5 SVM优缺点

SVM有如下主要几个特点:

(1)非线性映射是SVM方法的理论基础,SVM利用内积核函数代替向高维空间的非线性映射;
(2)对特征空间划分的最优超平面是SVM的目标,最大化分类边际的思想是SVM方法的核心;
(3)支持向量是SVM的训练结果,在SVM分类决策中起决定作用的是支持向量;
(4)SVM 是一种有坚实理论基础的新颖的小样本学习方法。
它基本上不涉及概率测度及大数定律等,因此不同于现有的统计方法。
从本质上看,它避开了从归纳到演绎的传统过程,实现了高效的从训练样本到预报样本的“转导推理”,
大大简化了通常的分类和回归等问题; (5)SVM 的最终决策函数只由少数的支持向量所确定,计算的复杂性取决于支持向量的数目,
而不是样本空间的维数,这在某种意义上避免了“维数灾难”。
(6)少数支持向量决定了最终结果,这不但可以帮助我们抓住关键样本、“剔除”大量冗余样本,
而且注定了该方法不但算法简单,而且具有较好的“鲁棒”性。
这种“鲁棒”性主要体现在:
①增、删非支持向量样本对模型没有影响; ②支持向量样本集具有一定的鲁棒性;
③有些成功的应用中,SVM 方法对核的选取不敏感 两个不足:

(1) SVM算法对大规模训练样本难以实施
由于SVM是借助二次规划来求解支持向量,
而求解二次规划将涉及m阶矩阵的计算(m为样本的个数),当m数目很大时该矩阵的存储和计算
将耗费大量的机器内存和运算时间。
针对以上问题的主要改进有
J.Platt的SMO算法、
T.Joachims的SVM、
C.J.C.Burges等的PCGC、
张学工的CSVM
以及O.L.Mangasarian等的SOR算法

(2) 用SVM解决多分类问题存在困难

经典的支持向量机算法只给出了二类分类的算法,
而在数据挖掘的实际应用中,一般要解决多类的分类问题。
可以通过多个二类支持向量机的组合来解决。
主要有
一对多组合模式、一对一组合模式和SVM决策树;
再就是通过构造多个分类器的组合来解决。
主要原理是克服SVM固有的缺点,结合其他算法的优势,解决多类问题的分类精度。
如:
与粗集理论结合,形成一种优势互补的多类问题的组合分类器。

6 总结

说明:上述流程中sacle 过程不一定要使用,具体问题具体分析。还是要多试验,多看效果。笔者在上述流程中犯了好多错误,希望读者尽量避免。

其实最难的部分在于特征信号的选取,即哪些因子作为feature去拟合label?

7 附录

官网gridregression.py下载地址 http://ntucsu.csie.ntu.edu.tw/~cjlin/libsvmtools/#grid_parameter_search_for_regression

(完)