摘要
主要研究了多类分类AdaBoost算法,及其在多类故障诊断问题中的应用。为了解决"一对一"算法和"一对余"算法的局限性,提出了基于决策树的AdaBoost算法。利用遗传算法的全局随机搜索性能对数据集进行特征筛选,得到新的特征数据集,根据CART算法构造决策树建立AdaBoost分类器,使得决策树每一个节点的最可分类别尽可能分开。通过对3个数据集进行仿真分析,表明该算法的性能优于其他2个算法,具有更高的通用性,验证了该算法的有效性。
Considering the limitations of the traditional methods,decision tree AdaBoost,which combines AdaBoost and decision tree,is proposed for multi-class classification.Genetic algorithm,as a global randomized search method,was used for selecting feature vectors from original dataset.The new feature vectors are got.Based on CART algorithm,decision tree is constructed,and AdaBoost classifiers are established,so the most separable classes could be separated at each node of decision tree.Compared with＂one-versus-one＂and＂one-versus-rest＂,simulations are conducted on three datasets to testify effectiveness.The results show that the proposed method has better performance and higher generalization ability than other two methods.
出处
《电子测量技术》
2011年第8期101-105,共5页
Electronic Measurement Technology
