Dissertation > Excellent graduate degree dissertation topics show

Research on Subspace Clustering Algorithm Based on Finding Effective Dimension

Author: LiuZhiGang
Tutor: RenJiaDong
School: Yanshan University
Course: Applied Computer Technology
Keywords: subspace clustering effective space effective dimension critical grid hash tree
CLC: TP311.13
Type: Master's thesis
Year: 2012
Downloads: 61
Quote: 0
Read: Download Dissertation

Abstract


In the field of data mining, subspace clustering is an important research hotspot.It has been widely used in many areas. Traditional clustering algorithms clusterdirectly in the original data space, but because of the sparse of the data in the originaldata space, it can not obtain the desired clustering results. To solve the above defect,this paper will focus on the research of the subspace clustering algorithm based onfinding effective space and subspace clustering algorithm based detecting critical gridfor data stream.First, we analyze the requirements of subspace clustering and the techniques fordimension reduction in high-dimensional data, at last, we analyze the classic subspaceclustering algorithms, including the top-down clustering algorithms and bottom-upclustering algorithms, and point out the advantages and disadvantages of thealgorithms.Second, in order to solve the problem that the algorithms based on grid discardthe sparse grids which are part of a cluster, we propose a subspace clustering algorithmbased on finding effective spaces. First, the algorithm finds the effective dimensionsaccording to relative entropy in every dimension. Then, it gets the high-dimensionaleffective spaces through a bottom-up iterative process, in the process, it uses themonotonicity of the clustering criterion to prune candidates, in addition, it also uses apruning function based on undirected acyclic connected graph. At last, it uses aclustering algorithm based on density to identify clusters in the effective spaces.Once more, in order to solve the problem that some clusters share one grid due tothe inappropriate partition of grids, we propose a subspace clustering algorithm basedon detecting critical grids for data stream. The algorithm is composed of the onlinecomponent and the offline component. In the online component, first, it finds theeffective dimensions according to relative entropy in every dimension. Then itgenerates the subspace by the correlation among dimensions and bit vector. At last, itprojects the data into grids and updates the eigenvector of the grids. In the offline component, when clustering request arrives, it computes adaptively the threshold ofthe dense grid, identifies the critical grids according to the centrality and attraction, atlast generates the clustering results.Finally, we implement the above algorithms with java language, and use syntheticdatasets and real datasets to test the clustering quality and scalability of our algorithms,furthermore, we analyse and compare the experimental results.

Related Dissertations

  1. Bayesian framework of image saliency detection,TP391.41
  2. Research on Density-Based Subspace Clustering Algorithm for Data Streams,TP311.13
  3. ESCHCD: Entropy-based Algorithm for Subspace Clustering with High Dimensional Categorical Datasets,TP311.13
  4. Research on Clustering Algorithms for High-Dimensional Data,TP311.13
  5. Research on Clustering Ensemble Methods and Their Applications,TP311.13
  6. Design and Implementation of Cluster Analysis Based on CMBG,TP18
  7. Collaborative Filtering in Personalized Recommendation Based on High-dimensional Subspace Clustering,TP393.09
  8. The Research on the Algorithm of Mining Quantitative Association Rules,TP311.13
  9. Research on Enhanced Soft Subspace Clustering Technology,TP311.13
  10. The Design and Analysis of Clustering Algorithms on Gene Expression Data,TP301.6
  11. The Research and Application of Subspace Clustering Algorithms,TP311.13
  12. Research on Subspace Clustering Algorithms Based on Density,TP311.13
  13. Semi-supervised Subspace Clustering Based on Space-level Constraint,TP18
  14. Research on Improved Subspace Clustering Algorithm,TP311.13
  15. Research on Subspace Clustering Algorithm on High-dimensional Categorical Datasets,TP311.13
  16. Research on Algorithms of Subspace Clustering Based on Pattern Similarity,TP311.13
  17. A Study of the Pattern-Based Clustering Theories,TP18
  18. The Research on Subspace Clustering for High Dimensional Data,TP301.6
  19. Find Interesting Frequent Patterns and Subspaces Using Length-Decreasing Support,TP311.13
  20. On Robust and High-Dimensional Data Clustering,TP311.13
  21. A Study on New Methods of Clustering and Image Thresholding in Unsupervised Learning,TP391.41

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