摘要
针对传统二叉树在多分类问题上存在分类精度不够高和时间复杂度较高的不足,提出了一种基于二叉树结构双优化的SVM多分类学习算法。此算法利用遗传算法对已经提取的特征参数子集和核参数进行双重优化,以获得最优的主要特征参数,从而有效地解决了样本结构复杂、分布不平坦的多分类识别问题。作者运用UCI数据库中的数据,通过仿真实验,并就经度和时间复杂度与有向无环图法和一对一法作比较,结果表明本文提出的算法具有较好的优越性。
Because of classification accuracy of the traditional binary tree for multi-classification problems is not high and it is too high for the time complexity, the authors of this paper present a new double optimization learning algorithm, based on the binary tree structure, which is a multi-classification algorithm. It makes the best of genetic algorithm to make feature parameters subset and kernel parameters optimized, in order to acquire the best important characteristic parameter combination for the purpose, and it can effectively solve the program of identification of complicated structure and uneven distribution sample. Combining with the UCI data in a database, through the simulation experiment, and compare the accuracy and time complexity with directed un-acyclic graph and one-to-one method, and the results show that the algorithm which has been proposed by the authors is effective in this paper.
出处
《重庆师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第6期109-113,共5页
Journal of Chongqing Normal University:Natural Science
基金
重庆师范大学博士研究基金(No.11XLB047)
关键词
GA
SVM
二叉树
多分类识别
GA
SVM
binary tree
multi-classification identification
