A Cluster Separation Measure

By David L. Davies, D. Bouldin. 1979

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Evaluation & Benchmarks

Citations in the corpus

Toward Domain-Invariant Speech Recognition via Large Scale Training (2018, methodology): “We also compute a cluster similarity measure R ij [32]: where S i is a dispersion measure of cluster i and M ij is a distance measure between clusters i and j .”
A Closer Look at Few-shot Classification (2019, methodology): “We use the Davies-Bouldin index Davies & Bouldin (1979) to measure intra-class variation.”
Scaling Multi-Agent Reinforcement Learning with Selective Parameter Sharing (2021, methodology): “The elbow method (Thorndike, 1953), the silhouette method (Rousseeuw, 1987), or the Davies–Bouldin index (Davies & Bouldin, 1979), all could be used to determine the number of clusters since our method tends to produce well-separated values.”
Privacy via the Johnson-Lindenstrauss Transform (2012, methodology): “, k-means++ clustering algorithm, scored according to the Davies-Bouldin index [14]) and records are sampled from a known distribution (e.”
Granular Computing: Perspectives and Challenges (2013, methodology): “Distance measures are used in cluster analysis to determine whether or not objects belong to a cluster [17], [31].”
Learning to Discover Novel Visual Categories via Deep Transfer Clustering (2019, background): “There are several CVI metrics, such as Silhou-ette [26], Dunn [13], DaviesBouldin [10], and Calinski-Harabasz [5]; while no metric is uniformly the best, the Sil-houette index generally works well [2, 3], and we found it to be a good choice for our case too.”
A Maximum Variance Cluster Algorithm (2002, methodology): “8 instance in [6], [ 8 ], [11], [16]. That is, the cluster algorithm may not be able to find the clustering corresponding to the local minimum or knee in the cluster tendency function 2.”
Cross-Validation for Imbalanced Datasets: Avoiding Overoptimistic and Overfitting Approaches [Research Frontier] (2018, methodology): “In this work, to find the optimal k number of clusters for each class, we have used three evaluation criteria: Calinski-Harabasz [42], Davies-Bouldin [43] and Silhouette [44], and a range of , .”
Learning hierarchical spatio-temporal pattern for human activity prediction (2016, methodology): “Therefore, Tnal can be scattered in plots named with English alphabets referred as the labels of actionlets according to the Davies–Bouldin index value [10], which is a good candidate for map unit clusters.”
A new cluster validity measure and its application to image compression (2004, unknown): “Davies–Bouldin’s measure [31], Gath–Geva’s measure”
You-Do, I-Learn: Discovering Task Relevant Objects and their Modes of Interaction from Multi-User Egocentric Video (2014, methodology): “We propose estimating the optimal number of clusters κ̂ using the standard Davies-Bouldin (DB) index [7].”
Performance Evaluation of a Novel iTongue for Indian Black Tea Discrimination (2010, background): “Davis–Bouldin [20] presented a cluster separation measure which represents class separability by the ratio of the sum of intracluster scatter to intercluster separation.”