改进多分类决策策略的RVM及其在液压泵故障诊断中的应用

发布时间：2018-03-22 19:46

本文选题：相关向量机　切入点：决策策略　出处：《计算机应用研究》2017年12期 　论文类型：期刊论文

【摘要】：针对传统多分类相关向量机(relevance vector machine,RVM)采用"最大票数赢(MVW)"决策策略的不足,为了提升相关向量机的多分类能力,首先改进了RVM的多分类决策策略,并利用具有Lévy飞行特征的果蝇算法(LFOA)对RVM核参数进行寻优,建立了LFOA-RVM分类模型。在适应度函数的评判下,果蝇种群经过多次迭代对指定范围内的核参数进行全局搜索寻优,完成模型建立。四组UCI标准数据集的MATLAB仿真实验结果表明,改进后的多分类决策策略和优化方法有效、可靠,能够提升RVM的分类能力;进一步将此模型应用于液压泵故障诊断,同样取得了较好的分类效果,验证了分类模型的有效性。
[Abstract]:In order to improve the ability of multi-classification vector machines, the multi-classification decision strategy of RVM is improved in order to improve the ability of multi-classification, aiming at the shortage of traditional multi-classification correlation vector machine (RVM) decision making strategy of "maximum votes win MVW", in order to improve the ability of multi-classification of correlation-vector machines, this paper first improves the multi-classification decision strategy of RVM. The RVM kernel parameters are optimized by using the fruit fly algorithm with L 茅 vy flight characteristics, and the LFOA-RVM classification model is established. Under the evaluation of fitness function, the fruit fly population performs global search and optimization of the kernel parameters in the specified range after repeated iterations. The MATLAB simulation results of four groups of UCI standard data sets show that the improved multi-classification decision strategy and optimization method are effective, reliable and can improve the classification ability of RVM, and the model is further applied to hydraulic pump fault diagnosis. A good classification effect is also obtained, and the validity of the classification model is verified.
【作者单位】：军械工程学院火炮工程系;
【基金】：河北省自然科学基金资助项目(E2016506003)
【分类号】：TH137.51;TP18

【相似文献】