Dissertation > Excellent graduate degree dissertation topics show

Research on Enhanced Soft Subspace Clustering Technology

Author: GuanQing
Tutor: WangShiTong
School: Jiangnan University
Course: Applied Computer Technology
Keywords: Cluster analysis Fuzzy Clustering Subspace clustering Feature Weighting Class separation
CLC: TP311.13
Type: Master's thesis
Year: 2011
Downloads: 68
Quote: 0
Read: Download Dissertation

Abstract

Cluster analysis is one of the key technologies in the field of data mining in e-commerce, information filtering, bioinformatics, pattern recognition and other fields are widely used. With more and more widespread application of clustering in practice, gradually highlights some of the problems, especially when dealing with large-scale, high-dimensional data was particularly evident. Currently, the cluster analysis of high-dimensional data is hot and difficult for the current study. In order to solve the problem of high-dimensional data clustering, R. Agrawal, first proposed the concept of subspace clustering. Be summed up in the subspace clustering divided into two categories: hard subspace clustering and soft subspace clustering. Hard subspace clustering method can identify the exact subspace where the different classes. Type hard subspace clustering is soft subspace clustering does not require precise sub-space for each class to find, but different weights given to each class characterized using these weights to measure each dimension characterized in that the different types of the contribution, ie, soft subspace clustering for each class to find a fuzzy subspace. This thesis in many soft subspace clustering algorithm, such algorithm exists an obvious shortcomings, for example, that almost all of the soft subspace clustering algorithm introduced into the class, such as tightness in the class , to construct the objective function. However, it can be expected to integrate into more discriminant information to construct a subspace clustering algorithm, the clustering performance will be further enhanced. Therefore, this article discusses the enhanced soft subspace clustering technology. The main work of this paper include the following aspects: the first part is an introduction section, a brief introduction of clustering techniques, current research and applications. The second part introduces the research background of the high-dimensional data clustering problem and solutions, with an emphasis on the subspace clustering algorithm, as well as commonly used three-seed space clustering algorithm. The third section describes the two types of typical soft subspace clustering algorithm: Fuzzy weighted subspace clustering and entropy weighted subspace clustering. The fourth part of the traditional fuzzy weighted soft subspace clustering using only the lack of information in the class, through the introduction of class identification information, enhanced fuzzy weighted soft sub-space clustering algorithm (EFWSSC). The new method first proposed combination of tightness separation between classes and class information to construct a new objective function in the fuzzy space, and then derive the new clustering rules and propose a new algorithm. Theoretical analysis and based on different data sets experimental results show that the new algorithm shows good effectiveness, better than most existing fuzzy weighted subspace clustering algorithm. The fifth part for the possibility of clustering algorithm (PCM) in the lack of high-dimensional data clustering, the introduction of subspace clustering, and presents the possibility of subspace clustering algorithm (SPC). SPC not only retains the PCM method has the advantage, but also has the advantages of the classic subspace clustering technology, i.e. better adaptability is displayed on the high-dimensional data, and can efficiently detect various types of which the sub-space. , The validity of the SPC and compared to the advantages of PCM algorithm has been verified through simulation experiments show that the simulated data sets and UCI data sets.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer software > Program design,software engineering > Programming > Database theory and systems
© 2012 www.DissertationTopic.Net Mobile