构造型三支决策模型优化及其在脑电信号中的应用研究

发布时间：2018-05-23 11:25

本文选题：三支决策模型 + 最近均值　；参考：《安徽大学》2017年硕士论文

【摘要】：三支决策理论符合人们做决策时的一贯行为。它具有三种决策规则,即接受、拒绝和不承诺。三支决策理论自提出以来,就被应用于不确定、不完整信息的分析判别中。在医疗诊断中,三支决策模型具有较好的应用前景。在研究粗糙集理论时诞生了三支决策模型,该模型最开始是为了合理的解释粗糙集的三个域(正域、负域和边界域)。基于DTRS的三支决策模型能够给出很好的决策结果,但是损失函数的选取带有人为主观的因素,不能很好的反映决策数据的客观信息。张对三支决策模型进行了拓展,将构造性覆盖算法引入到三支决策理论中,解决了获取阈值的问题。但是,经过研究发现虽然基于CCA的三支决策模型解决了主观获取的阈值问题,然而,该模型在形成覆盖的过程中,得到的覆盖中心是随机选取的。考虑到被分类的数据一般都具有一定的规律性,即同类别的数据在空间分布上总是相近的,而随机选取覆盖中心未必能够形成最优覆盖。所以,本文首先对随机选取覆盖中心的三支决策模型进行优化。然后,将优化后的三支决策模型引入到帕金森症患者的脑电信号的分析判别研究中。优化覆盖中心的过程中,主要用到最近均值的思想,选择最佳的覆盖中心。将优化覆盖中心的三支决策模型应用到脑电信号的分析判别处理上是本文的一大亮点。传统的二分类模型在脑电信号分类判别中并不符合医生在诊断上的原则,而结合了离散小波变换、样本熵以及基于优化覆盖中心的三支决策模型的脑电信号分析模型能够在临床诊断中给出准确的判断,为医生的诊断提供重要的依据。本文从三支决策模型的理论和应用两个方面进行研究,主要工作包括:1.本文首先对三支决策的提出、发展及应用进行了介绍,详细阐述了三支决策模型。对脑电信号分析方面的知识和研究进行了梳理,详细描述了脑电信号分析的研究现状和进展。并主要对基于CCA的三支决策模型进行了重点研究。之后,针对基于CCA的三支决策模型覆盖中心随机选取的问题结合最近均值思想提出了一种优化覆盖中心的三支决策模型。2.本文提出的优化覆盖中心的三支决策模型,解决了随机选取覆盖中心的问题,并且提高了分类的正确率,效果更好。而在脑电信号分析判别中,二分类模型应用广泛,但是在数据不充足时,此种模型普遍不能提供让患者进一步诊断的的信息,这种在临床上,不能根据当前信息立即做出诊断是否患病的情况是真实存在的,即延迟判断,需要提供更多信息做出判断。这种情况的存在为三支决策模型应用到脑电信号分析中提供了依据。3.三支决策理论的三个域(接受、拒绝、不承诺)在临床诊断中可对应于患病、未患病、延迟判断,即需要更多信息进行判断。本文将基于O_CCA的三支决策模型应用到帕金森症脑电信号的分析判别研究中。提出了一种结合离散小波变换、样本熵和基于O_CCA的三支决策模型的脑电信号分析判别方案。本文在帕金森症脑电信号数据上针对离散小波变换去除噪声能力以及脑电信号分析中常用二分类模型进行了实验,结果表明本方案能够获取更好的分析效果。
[Abstract]:The three decision theory conforms to the consistent behavior of people in making decisions. It has three kinds of decision-making rules, namely, acceptance, rejection and non commitment. Since the three decision theory has been proposed, it has been applied to the analysis and discrimination of uncertainty and incomplete information. In the medical diagnosis, the three decision-making models have a good application prospect. In the study of rough set theory, the theory of rough sets is studied. Three decision models were born, the first of which was to reasonably explain the three domains of rough sets (positive domain, negative domain and boundary domain). The three decision models based on DTRS can give good decision results. But the selection of the loss function has the subjective factors, which can not reflect the objective information of the decision data well. The three decision models are extended, and the constructive coverage algorithm is introduced into three decision theories to solve the problem of obtaining the threshold. However, it is found that although the three decision models based on CCA have solved the threshold problem of subjective acquisition, the coverage center of the model is randomly selected in the process of forming and covering. Considering that the classified data generally have some regularity, that is, the data of the same category are always similar in the spatial distribution, and the random selection of the coverage center may not be able to form the optimal coverage. Therefore, this paper first optimizes the three decision models which randomly select the coverage center. Then, the three decision models are optimized. In the analysis and discrimination of EEG signals of patients with Parkinson's disease, in the process of optimizing the coverage center, the idea of the nearest mean is used and the best coverage center is chosen. The three decision model of the optimized coverage center is applied to the analysis and discrimination of the EEG signal. The traditional two classification model is in this paper. The classification and discrimination of EEG does not conform to the doctor's principle of diagnosis, but it combines the discrete wavelet transform, the sample entropy and the three decision model based on the optimal coverage center. The EEG signal analysis model can give the accurate judgment in the clinical diagnosis and provide the important basis for the doctor's diagnosis. This paper from three decision models Two aspects of the theory and application are studied. The main work is as follows: 1. this article first introduces the development and application of three decisions, expounds the three decision models in detail, combs the knowledge and research of the EEG analysis, and describes the research status and progress of the EEG analysis in detail. Three decision models based on CCA are focused on. Then, in view of the random selection of three decision models based on the three decision models based on the nearest mean idea, a three decision model,.2., is proposed to optimize the coverage center, and the three decision model of the optimal coverage center is proposed, which solves the random selection center. The two classification model is widely used in EEG analysis and discrimination, but in the case of insufficient data, this model generally does not provide information for further diagnosis. In clinical, it is not possible to diagnose the disease immediately according to the current information. The existence of the real existence, that is, the delay judgment, needs to provide more information to make a judgment. The existence of this situation provides three decision models to the EEG analysis by providing three domains (acceptance, rejection, non commitment) in three domains (acceptance, rejection, non commitment) in the application of three decision models to the EEG analysis. In this paper, three decision models based on O_CCA are applied to the analysis and discrimination of Parkinson's EEG signal. A discriminant scheme of EEG signal analysis is proposed, which combines discrete wavelet transform, sample entropy and three decision model based on O_CCA. This paper is used to remove the discrete wavelet transform on the Parkinson's EEG data. Experiments on noise classification and two classification models in EEG analysis show that the scheme can achieve better results.
【学位授予单位】：安徽大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R318;TN911.6

【相似文献】