基于免疫遗传算法的支持向量机参数优化及其应用

Parameter Optimization and It's Application of Support Vector Machine Based on IGA

支持向量机是在统计学习理论基础上发展起来的一种性能优良的新型机器学习方法,它具有坚实的理论基础,巧妙的算法实现。支持向量机的卓越性能依赖于它的参数的正确选择。本文采用改进的免疫遗传算法对支持向量机的参数进行优化。实验证明对于低维数据分类时,本文的优化算法比传统的网格法可以较大减少参数优化时间和提升分类的准确率。对高维的文本数据分类时,在保证分类准确率的前提下,仍然可以较大减少优化的时间。

Support Vector Machine（SVM） is developed on the frame of the statistical learning theory, which has been a new ex- cellent machine learning method. SVM has solid theoretical foundation, clever algorithms. The outstanding performance of SVM depends on the choice of model parameters, including penalty parameter and kernel function parameter. In the paper, an im- proved Immune Genetic Algorithm （IGA） is applied to optimize the parameter of SVM. The experiment result proves that the method the paper proposed can reduce optimal time and improve the precision of classification largely compared to Grid searching when the classifier object is low dimension. When the classifier object is high dimension such as text data, the method can also reduce ontimal time lareelv eomnared to Grid searehine in the case of keeoin~ same precision of classification.