A Cluster Separation Measure

Davies, David; Bouldin, Donald W.

doi:10.1109/tpami.1979.4766909

Cited by 6,834 publications

(3,271 citation statements)

References 2 publications

Supporting

Mentioning

3,199

Contrasting

Unclassified

Order By: Relevance

“…To inform this decision, a selection of statistics were used to choose the number of clusters that best represents the data. These included the 'C-index' (Hubert & Levin 1976) and Davies & Bouldin (1979) 'Validity Index' which both examine the similarity of clusters. These measures have both been shown to be effective at determining the correct number of clusters (Milligan & Cooper 1985).…”

Section: Resultsmentioning

confidence: 99%

A neighbourhood level mortality classification of England and Wales, 2006–2009

2014

View full text Add to dashboard Cite

The paper provides an overview of a neighbourhood level classification of mortality for England and Wales (2006-2009). Standardised mortality ratios for 63 causes of death were calculated for middle super output areas (weighted by prevalence). A k-means partitional method was used to classify the data. An eight cluster solution was found to best segment mortality patterns. Clusters mostly differentiated in terms of prevalence, however the importance of neurodegenerative diseases and causes related to unhealthy behaviours were important. The results describe a neighbourhood classification that can be an important tool to help inform policy development, resource allocation and targeting of services.

show abstract

Section: Resultsmentioning

confidence: 99%

A neighbourhood level mortality classification of England and Wales, 2006–2009

2014

View full text Add to dashboard Cite

show abstract

“…SOM units within each cluster are more similar to each other than they are to SOM units of other clusters. The Davies-Bouldin index was used to define the appropriate number of clusters for the k-means algorithm (Davies and Bouldin 1979). This index shows for which number of clusters the ratio of average within-to-between cluster distances is minimal, i.e., how many clusters are optimal for the dataset.…”

Section: Statisticsmentioning

confidence: 99%

Changes in balance coordination and transfer to an unlearned balance task after slackline training: a self-organizing map analysis

et al. 2017

View full text Add to dashboard Cite

effect size was more than three times larger in the slackline. The SOM analysis, followed by a k-means clustering and marginal homogeneity test, showed that the balance coordination pattern was significantly different between pre-and post-test for the slackline task only (χ 2 = 82.247; p < 0.001). The shift in balance coordination on the slackline could be characterized by an increase in range of motion and a decrease in velocity and frequency in nearly all degrees of freedom simultaneously. The observation of low transfer of coordination strategies to the flamingo task adds further evidence for the task-specificity principle of balance training, meaning that slackline training alone will be insufficient to increase postural control in other challenging situations.

show abstract

“…K-means clustering defines closeness as the metric for similarity to group data sets into clusters. With k-means, the similarity of data set is defined as their closeness, whereas the dissimilarity is determined as the separation of cluster centers in an Euclidian plane [16]. K-means aims to determine the best number of clusters for a given data set and to divide the data points into the corresponding K groups.…”

Section: B Nuclei Clustersmentioning

confidence: 99%

“…The data can be grouped by minimizing the total squared distances between data points and centroids (i.e. geometric mean of clusters) inside a cluster and by maximizing the separation between different clusters [8], [17], [16]. Several algorithms are reported in the literature for computation of the best number of clusters [6], [9], [18], [19].…”

Section: B Nuclei Clustersmentioning

confidence: 99%