Dissertation > Excellent graduate degree dissertation topics show

Density-based clustering algorithm

Author: SunLingYan
Tutor: YangMing
School: University of North
Course: Applied Mathematics
Keywords: Cluster analysis Fast Algorithm Core Point Represents the object Relative Density
CLC: TP301.6
Type: Master's thesis
Year: 2009
Downloads: 294
Quote: 3
Read: Download Dissertation

Abstract

With the development of information technology, data mining technology has been widespread concern. In data mining technology, there are many research areas, cluster analysis is an important research direction. And classification of different clustering aims without any prior knowledge of the premise, the data similarity aggregates the data into different clusters, so that the same elements in the cluster as similar as possible, the different elements in the cluster as the difference large, it is also known as unsupervised classification. Clustering analysis as a data mining system in one module, either as a separate tool to discover the underlying distribution of data in the database information can also be used for other data mining analysis algorithm, a preprocessing step, so studying how to improve clustering algorithm Performance has important significance. So far researchers have proposed a variety of clustering algorithms, such as division method, hierarchical method, grid-based approach, density-based methods, model-based methods. Since density-based clustering algorithm to discover clusters of arbitrary shape, identification dataset noise points, scalability and good features, in many areas has important applications. DBSCAN algorithm is a typical density clustering algorithm, but since the overall density algorithm, part of the cluster density data may be processed as a noise, and the edge of the point in the two clusters, the density of the point if there is a relatively large the situation is likely to cause a simply connected case, an error results. Meanwhile algorithms need to determine whether each point in the database as the core point, create a query for each point region, which will require frequent I / O operations. FDBSCAN algorithm is an improved algorithm DBSCAN algorithm. The algorithm used in the neighborhood part of the core point point as seed points to extend the cluster, thus greatly reducing the area of ??the number of queries, reduce I / O overhead to some extent accelerated the clustering speed. But its easy to lose part of the clustering process objects become noise, affecting the clustering results. Chapter III of this algorithm on FDBSCAN-depth study of the problems on the basis of specific proposed a core object choose the most distant way, and its core target for non-core point is not to make inquiries, so that the case is lost objects discussed in detail. Finally, from the core area of ??the core points chosen to represent the object's methods, to some extent solve the problem of the missing objects. FDBSCAN DBSCAN algorithm algorithm is an improvement in speed, based on the relative density clustering algorithm RDBClustering (Relative Density Based Clustering) algorithm is used for its overall density of this shortcoming to do improvements. Although the two algorithms from different angles on DBSCAN algorithm has been improved, but there is still insufficient. The former to some extent accelerated the clustering speed, but can not resolve objects uneven density error results when clustering problem; solve the overall density of the latter problem, but its speed is very slow, the memory required is relatively large. Therefore, this chapter on the basis of the two proposed a new algorithm - based on the relative densities of Fast clustering algorithm FRDBClustering (Fast Relative Density-Based Clustering), the new algorithm combines the FDBSCAN algorithm and RDBClustering two algorithms advantages, not only solved the problem of global parameters of DBSCAN, to a certain extent, also accelerated the clustering speed, the experiment proved the effectiveness of this method.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > General issues > Theories, methods > Algorithm Theory
© 2012 www.DissertationTopic.Net Mobile