基于模糊二叉树的支持向量机算法研究及应用

发布时间：2018-06-08 06:21

本文选题：支持向量机 + 模糊隶属度　；参考：《山东科技大学》2017年硕士论文

【摘要】：支持向量机自20世纪90年代由Vapnik提出以来,就以其卓越的学习能力成为了机器学习的研究热点。传统的支持向量机是研究两类分类的,而实际应用中多类分类更加常见,同时实际应用中会遇到大量模糊信息,如何将其更好的应用于多类分类领域以及克服模糊信息的干扰是支持向量机研究的重点。本文针对现有的模糊二叉树支持向量机在树形结构的构造和隶属度函数的构造方法上存在的不足进行了改进,其主要工作如下:1.针对模糊二叉树支持向量机现有的树形结构构造方法的缺点,提出了基于k-means聚类算法的构造方法,这是一种自下而上的树形结构的构造方法,通过对样本不断的聚类,使得每次不同类的结合更加合理,且对构造的树形结构不强求为完全正态的,更加符合实际情况,同时通过数值实验验证了这种方法的有效性。2.研究了模糊二叉树支持向量机的隶属度函数,提出了基于相关系数的隶属度函数构造方法,根据样本点的相关程度赋予不同的值,以此来实现样本的模糊化处理,从而避免了样本密度不均造成的影响,并通过数值实验验证了其可行性。3.提出了一种改进的模糊二叉树支持向量机,并将其与1-v-rSVMs和1-v-1SVMs作分类性能的比较实验,结果表明改进的模糊二叉树支持向量机总体性能高于1-v-rSVMs和1-v-1SVMs的性能。最后,对论文进行了总结,并对今后的工作做出了展望。
[Abstract]:Support vector machine (SVM) has been a hot topic in machine learning since it was proposed by Vapnik in 1990s because of its excellent learning ability. The traditional support vector machine studies two kinds of classification, but the multi-class classification is more common in the practical application, at the same time, it will encounter a lot of fuzzy information in the practical application. How to better apply it to the field of multi-class classification and overcome the interference of fuzzy information is the focus of support vector machine (SVM) research. In this paper, the shortcomings of the existing fuzzy binary tree support vector machine in the construction of tree structure and membership function are improved. The main work of this paper is as follows: 1. Aiming at the shortcoming of the existing tree structure construction method of fuzzy binary tree support vector machine, this paper proposes a construction method based on k-means clustering algorithm, which is a bottom-up tree structure construction method. The combination of different classes is more reasonable, and the tree structure is not required to be completely normal, which is more in line with the actual situation. At the same time, the validity of this method is verified by numerical experiments. In this paper, the membership function of fuzzy binary tree support vector machine is studied, and the method of constructing membership function based on correlation coefficient is put forward. According to the correlation degree of sample points, different values are assigned to realize the fuzzy processing of samples. In order to avoid the influence caused by the uneven density of the sample, the feasibility of the method is verified by numerical experiments. An improved fuzzy binary tree support vector machine is proposed and compared with 1-v-rSVMs and 1-v-1 SVMs. The results show that the overall performance of the improved fuzzy binary tree support vector machine is better than that of 1-v-rSVMs and 1-v-1 SVMs. Finally, the paper is summarized and the future work is prospected.
【学位授予单位】：山东科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【参考文献】