A Resampling Approach to Cluster Validation

Röth, Volker; Lange, Tilman; Braun, Mikio L.; Buhmann, Joachim M.

doi:10.1007/978-3-642-57489-4_13

Cited by 89 publications

(58 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…As was mentioned earlier, the methodology was employed in Lange et al (2002Lange et al ( , 2003Lange et al ( , 2004 and Breckenridge (1989). It offers to evaluate the likeness of clusterings via a classifier trained by means of a second (clustered) dataset.…”

Section: First Simulation Methodsmentioning

confidence: 99%

“…In the framework of our method and the other approaches testing the goodness of partitions for the possible number of clusters belonging to a given interval (see, for example, Roth et al 2002;Lange et al 2003Lange et al , 2004Walther 2005 andDudoit andFridlyand 2002) the tradeoff between the compression and the information rates can be expressed by the interval selection. Essentially, this area, [k min , k max ] in the algorithm presented in Sect.…”

Section: Comments Regarding the Algorithmmentioning

confidence: 99%

“…For example, this method has been implemented in the known Clest method of Dudoit and Fridlyand (2002). A general prediction based resampling approach has been proposed in Roth et al (2002), Lange et al (2003Lange et al ( , 2004. Tibshirani and Walther (2005) employ a similar prediction strength procedure.…”

Section: Introductionmentioning

confidence: 99%

“…So, the approaches presented in Lange et al (2002Lange et al ( , 2003Lange et al ( , 2004 and Tibshirani and Walther (2005) employ Breckenridge's idea (Breckenridge 1989). This approach assesses the cluster solutions resemblance, resting upon a classifier trained by means of a second (clustered) dataset.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Resampling approach for cluster model selection

et al. 2011

View full text Add to dashboard Cite

In cluster analysis, selecting the number of clusters is an "ill-posed" problem of crucial importance. In this paper we propose a re-sampling method for assessing cluster stability. Our model suggests that samples' occurrences in clusters can be considered as realizations of the same random variable in the case of the "true" number of clusters. Thus, similarity between different cluster solutions is measured by means of compound and simple probability metrics. Compound criteria result in validation rules employing the stability content of clusters. Simple probability metrics, in particular those based on kernels, provide more flexible geometrical criteria. We analyze several applications of probability metrics Editors: Süreyya Özögür-Akyüz, Devrim Ünay, and Alex Smola. Learn (2011) 85:209-248 combined with methods intended to simulate cluster occurrences. Numerical experiments are provided to demonstrate and compare the different metrics and simulation approaches.

show abstract

Section: First Simulation Methodsmentioning

confidence: 99%

Section: Comments Regarding the Algorithmmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Resampling approach for cluster model selection

et al. 2011

View full text Add to dashboard Cite

show abstract

“…Over the years, several approaches to this problem have been suggested. Among them, the more well-known ones include crossvalidation [18], penalized likelihood estimation [19,20], resampling [21], and finding the 'knee' of an error curve [2,22]. However, these techniques either make a strong parametric assumption, or are computationally expensive.…”

Section: Related Workmentioning

confidence: 99%

A Spectroscopy of Texts for Effective Clustering

Ong

et al. 2004

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. For many clustering algorithms, such as k-means, EM, and CLOPE, there is usually a requirement to set some parameters. Often, these parameters directly or indirectly control the number of clusters to return. In the presence of different data characteristics and analysis contexts, it is often difficult for the user to estimate the number of clusters in the data set. This is especially true in text collections such as Web documents, images or biological data. The fundamental question this paper addresses is: "How can we effectively estimate the natural number of clusters in a given text collection?". We propose to use spectral analysis, which analyzes the eigenvalues (not eigenvectors) of the collection, as the solution to the above. We first present the relationship between a text collection and its underlying spectra. We then show how the answer to this question enhances the clustering process. Finally, we conclude with empirical results and related work.

show abstract