Rank: 326 based on 364 downloads (last 30 days) and 7 files submitted
photo

Kaijun Wang

E-mail
Company/University
Fujian Normal University

Personal Profile:
Professional Interests:

 

Watch this Author's files

 

Files Posted by Kaijun View all
Updated   File Tags Downloads
(last 30 days)
Comments Rating
06 Aug 2012 Screenshot Recognizing Far-Near Relations of Clusters by GDEM, Visualization by Line-Pearl Pattern measure far-near degrees (distances) between clusters & dense degrees of border regions of clusters Author: Kaijun Wang statistics, cluster analysis 12 1
  • 5.0
5.0 | 1 rating
21 Aug 2010 Fast Affinity Propagation Clustering under Given Number of Clusters Fast searching the given number of clusters Author: Kaijun Wang clustering, statistics, fast ap clustering 32 0
26 Jul 2009 Adaptive Affinity Propagation clustering advantage of speed & performance appears under large number of clusters & large dataset Author: Kaijun Wang statistics, probability, affinity propagation, clustering, large number of clust..., clustering algorithm 68 6
  • 4.4
4.4 | 5 ratings
25 Jul 2009 CVAP: Cluster Validity Analysis Platform (cluster analysis and validation tool) supplying over 17 validity indices and 5 clustering algorithms based on GUI Author: Kaijun Wang statistics, probability, clustering analysis, validity index, cluster validation 77 18
  • 4.36364
4.4 | 11 ratings
08 Jul 2009 (simple) Tool for estimating the number of clusters 12 validity indices, illustrate estimation of the number of clusters Author: Kaijun Wang statistics, probability, validity index, cluster validation, clustering analysis, number of clusters 123 5
  • 4.25
4.2 | 4 ratings
Comments and Ratings on Kaijun's Files View all
Updated File Comment by Comments Rating
18 Dec 2013 CVAP: Cluster Validity Analysis Platform (cluster analysis and validation tool) supplying over 17 validity indices and 5 clustering algorithms based on GUI Author: Kaijun Wang P, NR

Please review valid_internal_deviation.m: I believe it has errors in the correction factors for Calinski-Harabasz and Krzanowski-Lai indices, and Davies-Bouldin index estimante apper to derive for incomplete pairwise estimates.

18 Dec 2013 CVAP: Cluster Validity Analysis Platform (cluster analysis and validation tool) supplying over 17 validity indices and 5 clustering algorithms based on GUI Author: Kaijun Wang P, NR

07 Nov 2013 (simple) Tool for estimating the number of clusters 12 validity indices, illustrate estimation of the number of clusters Author: Kaijun Wang Dernoncourt, Franck

It would be great to add the ruspini.mat dataset in the submission. (currently available at https://code.google.com/p/a2hbc/source/browse/trunk/a2hbc/scripts/common/LIBRA/ruspini.mat)

03 Sep 2013 CVAP: Cluster Validity Analysis Platform (cluster analysis and validation tool) supplying over 17 validity indices and 5 clustering algorithms based on GUI Author: Kaijun Wang Chenghao

When applying kmeans algorithm with euclidean distance function, kmeans algorithm in valid_clusteringAlgs.m (at line 29 of CVAP 3.7 version) uses the distance function R = 'sqEuclidean';. However, when evaluating with silhouette measure in valid_internal.m(at line of 28), the code uses R = 'euclidean';. I think the same distance function should be used.

11 Jun 2013 CVAP: Cluster Validity Analysis Platform (cluster analysis and validation tool) supplying over 17 validity indices and 5 clustering algorithms based on GUI Author: Kaijun Wang Ilc, Nejc

Many thanks to the author for his valuable and useful contribution. However, I think there is a bug in the implementation of the Dunn's index.

I have compared results from CVAP with the Julian Ramos' implementation (http://www.mathworks.com/matlabcentral/fileexchange/27859) and also with R package clValid. Output values from clValid and Ramos' code are identical, whereas the CVAP results are not in an agreement with them. I think there are errors in computing the diameter of clusters and the shortest distance between clusters (function 'valid_sumsqures'). Please, consider revising this part of code.

Contact us