粗糙集—决策树在故障诊断中的应用研究

发布时间：2018-01-08 12:13

本文关键词：粗糙集—决策树在故障诊断中的应用研究　出处：《东北大学》2014年硕士论文　论文类型：学位论文

【摘要】：目前,把多种数据挖掘技术相结合从而实现对故障问题的诊断已经成为未来故障诊断研究的发展趋势。决策树模型凭借其可读性强、分类速度快等优点在故障诊断领域发挥着不可替代的作用,然而,训练数据集中的噪声数据以及在建模过程中存在的过拟合问题严重制约着决策树模型的诊断效率。粗糙集相关理论则能在保持数据分类能力不变的前提下,提高模型对噪声数据的容忍程度,从而扩大对数据的应用范围。故二者的结合成为故障诊断领域新的研究方向。本文以构建粗糙集-决策树模型以及实现对故障数据的诊断为目的,主要做了以下几方面的工作：首先,对粗糙集的经典理论和决策树算法的进行了梳理,分别分析了这两种方法与其他数据挖掘手段结合下的应用现状。其次,在比较3种决策树算法后,提出用C4.5算法替代前人提出的粗糙集-决策树算法中的ID3算法,从而使模型能够克服因属性取值不同带来的误差。再次,针对已有的粗糙集-决策树模型不能很好的克服噪声数据这一现象,引入粗糙集理论中的变精度概念,将其应用到决策树初始变量的选择过程中,从而实现改进后的决策树模型能够克服一定程度下的噪声数据。最后,把改进后的粗糙集—决策树模型应用到故障诊断数据中,并与C4.5决策树模型进行有关比较,验证了前者相较于后者,决策树规模更小,预测能力更好,生成的规则更加丰富。
[Abstract]:At present, various data mining technology combined to realize the fault diagnosis problem has become the future development trend of fault diagnosis. The decision tree model with its readability, classification speed and other advantages in the field of fault diagnosis plays an irreplaceable role, however, the noise data and existing in the training data set in the process of modeling the over fitting problem restricts the efficiency of diagnosis decision tree model. Rough set theory can while keeping the classification ability of data at the same time, improve the degree of tolerance to noise data model, so as to expand the scope of application of data binding. So the two is becoming a new research direction in the field of fault diagnosis. This paper focuses on the construction of rough set and decision tree model and realize the fault diagnosis data for the purpose, mainly do the following work: firstly, the classical rough sets theory Theory and decision tree algorithm are summarized, analyzed the two methods and other methods of data mining combined with the application situation. Secondly, after comparing 3 kinds of decision tree algorithms, proposed previously proposed rough set with C4.5 algorithm of decision tree algorithm ID3 algorithm, which can overcome the model because the error caused by the different attribute values. Thirdly, according to the existing rough set and decision tree model can well overcome the phenomenon of noise data, using rough set of variable precision concept in theory, which is used to select the initial decision tree variables, so as to realize the improved decision tree model can overcome the noise the data under a certain degree. Finally, the improved rough set and decision tree model is applied to the fault diagnosis data, and the comparison with the C4.5 decision tree model, verified the former compared with the latter, the decision tree size Smaller, better predictive, and more abundant rules.

【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：C934

【参考文献】