基于混合核函数的SVM及其应用研究

发布时间：2018-05-21 02:40

本文选题：支持向量机 + 核函数　；参考：《大连海事大学》2016年硕士论文

【摘要】：支持向量机(SVM)是由Vapnik提出的基于统计学习理论的一种分类方法,使低维特征空间经过一种非线性转换得到一个高维特征空间,最优分类超平面即可在转变到这个新的空间中求解,由此将不可分问题转化为可分问题。在解决类似小样本、高维数、非线性等众多实际问题中SVM表现了很强的优势,这种方法在机器学习中占有着非常重要的地位。核函数作为支持向量机的核心,支持向量机的性能表现受到不同的核函数的直接影响,提高核函数性能成为研究支持向量机工作中的核心问题之一。本文首先介绍了支持向量机理论和核函数理论,通过对SVM及核函数性质的研究并在不同数据集上测试四种常用核函数(线性核函数、高斯核函数、多项式核函数、sigmoid核函数)的性能,选择较优的两种或多种核函数进行组合构造新的混合核函数,然后利用遗传算法这种人工智能学习算法对混合核函数中核权重系数及核参数进行自动化优化,同时测试混合核函数SVM在不同数据集上的性能,并与单核测试数据结果进行分析比较,得出结论,验证混合核函数支持向量机的实用性。在文章最后对全文研究工作进行了概况总结,需要研究探讨的问题也得到了进一步的提出。
[Abstract]:Support vector machine (SVM) is a classification method based on statistical learning theory proposed by Vapnik. The optimal classification hyperplane can be solved in this new space, thus the inseparability problem can be transformed into a separable problem. SVM has a strong advantage in solving many practical problems such as small sample, high dimension, nonlinear and so on. This method plays a very important role in machine learning. Kernel function is the core of support vector machine. The performance of support vector machine is directly affected by different kernel functions. Improving the performance of kernel function has become one of the core problems in the work of support vector machine. In this paper, the support vector machine theory and kernel function theory are introduced. By studying the properties of SVM and kernel function, four common kernel functions (linear kernel function, Gao Si kernel function) are tested on different data sets. The performance of polynomial kernel function sigmoid kernel function, two or more kernel functions are selected to combine to construct a new hybrid kernel function. Then the artificial intelligence learning algorithm is used to automate the optimization of kernel weight coefficient and kernel parameter in hybrid kernel function, and the performance of hybrid kernel function SVM on different data sets is tested at the same time. The results are analyzed and compared with the data of single core test, and the results show that the hybrid kernel support vector machine is practical. At the end of the paper, the research work is summarized, and the problems that need to be studied are also put forward.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP18

【相似文献】