不完备信息系统上的快速属性约简算法

发布时间：2018-05-06 02:39

本文选题：不完备信息系统 + 属性约简　；参考：《大连海事大学》2017年硕士论文

【摘要】：在粗糙集的理论研究中,有很多重要的概念,其中的属性约简是在粗糙集研究中的最重要的研究内容。随着数据量的增加,在粗糙集模型上进行分类效率大幅度的降低,因此属性约简显得尤为重要。在现实生活中,数据的完整性无法得到保证,因此在实际应用中的数据大多是不完整的。由此不完备的信息系统应运而生,但是传统的粗糙集模型无法直接处理不完备信息,因此不完备信息系统成为粗糙集模型研究上的重要的方向。目前已有的处理不完备信息系统的方法主要有基于容差关系和基于差别矩阵的。这两种方法的时间效率很低,因此本文通过对已有的属性约简算法的研究提出一种不完备信息系统上的快速属性约简算法。本文提出的不完备信息系统上的快速属性约简算法,将邻域粗糙集引入不完备信息系统,实现了基于属性隶属度的属性约简,并且通过将桶引入到不完备信息系统上,大大的提高了属性约简的效率,同时保证了属性约简之后的数据集的分类精度。基于扩展的邻域粗糙集模型的不完备信息系统上的快速属性约简会根据邻域阈值的不同产生不同的属性约简的结果。而邻域阈值的设置都是人为设置缺乏理论依据,因此本文提出基于最大近邻的粗糙集模型上的属性约简算法。数据本身出发得到最大近邻来代替邻域阈值,更具有理论依据,提高了属性约简后数据集的分类精度。
[Abstract]:In the theoretical research of rough sets, there are many important concepts, among which attribute reduction is the most important research content in the research of rough sets. With the increase of data volume, the classification efficiency of rough set model is greatly reduced, so attribute reduction is particularly important. In real life, the integrity of data can not be guaranteed, so most of the data in practical applications are incomplete. The incomplete information system emerges as the times require, but the traditional rough set model can not deal with the incomplete information directly, so the incomplete information system becomes an important direction in the research of the rough set model. The existing methods to deal with incomplete information systems are mainly based on tolerance relation and discriminant matrix. The time efficiency of these two methods is very low, so this paper proposes a fast attribute reduction algorithm for incomplete information systems by studying the existing attribute reduction algorithms. In this paper, the fast attribute reduction algorithm for incomplete information system is proposed. The neighborhood rough set is introduced into the incomplete information system. The attribute reduction based on attribute membership degree is realized, and the bucket is introduced to the incomplete information system. The efficiency of attribute reduction is greatly improved, and the classification accuracy of data set after attribute reduction is ensured. Fast attribute reduction on incomplete information systems based on extended neighborhood rough set model will result in different attribute reduction according to the threshold of neighborhood. However, there is no theoretical basis for the setting of neighborhood threshold. Therefore, an attribute reduction algorithm based on the maximum nearest neighbor rough set model is proposed in this paper. Starting from the data itself, the maximum nearest neighbor is obtained to replace the neighborhood threshold, which has more theoretical basis and improves the classification accuracy of the data set after attribute reduction.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【相似文献】