Finding the Number of Clusters in a Dataset

Sugar, Catherine A.; James, Gareth

doi:10.1198/016214503000000666

Cited by 735 publications

(506 citation statements)

References 3 publications

Supporting

Mentioning

496

Contrasting

Unclassified

Order By: Relevance

“…As seen in Table 5, no correlation was found between dataset length and the optimum number of the clusters, which well suits the literature findings [41,42]. The optimum number of the clusters may vary based on the properties of the dataset, such as the geometric distribution, statistical measures, and neighborhood measures [42,48].…”

Section: Batch Datasetssupporting

confidence: 79%

“…This is achieved by determining the optimum cluster centers, which minimize the distortion of the samples in the clusters [41]. The process of optimizing cluster centers and members is done iteratively such that k samples are selected arbitrarily as cluster centers and the cluster centers are updated until reaching optimum centers [6,17].…”

Section: K-means Clusteringmentioning

confidence: 99%

See 1 more Smart Citation

A novel approach for extracting ideal exemplars by clustering for massivetime-ordered datasets

Ertuğrul¹

2017

Turk J Elec Eng & Comp Sci

View full text Add to dashboard Cite

Abstract:The number and length of massive datasets have increased day by day and this yields more complex machine learning stages due to the high computational costs. To decrease the computational cost many methods were proposed in the literature such as data condensing, feature selection, and filtering. Although clustering methods are generally employed to divide samples into groups, another way of data condensing is by determining ideal exemplars (or prototypes), which can be used instead of the whole dataset. In this study, first the efficiency of traditional data condensing by clustering approach was confirmed according to obtained accuracies and condensing ratios in 9 different synthetic or real batch datasets. This approach was then improved to be employed in time-ordered datasets. In order to validate the proposed approach, 23 different real time-ordered datasets were used in experiments. Achieved mean RMSEs were 0.27 and 0.29 by employing the condensed (mean condensed ratio was 97.17%) and the whole datasets, respectively. Obtained results showed that higher accuracy rates and condensing ratios were achieved by the proposed approach.

show abstract

Section: Batch Datasetssupporting

confidence: 79%

Section: K-means Clusteringmentioning

confidence: 99%

A novel approach for extracting ideal exemplars by clustering for massivetime-ordered datasets

Ertuğrul¹

2017

Turk J Elec Eng & Comp Sci

View full text Add to dashboard Cite

show abstract

“…The number of clusters is selected by testing an hierarchical cluster analysis using Ward's method in order to obtain a general view of the clustering step by step in the corresponding dendrogram. Also, a technique introduced by Sugar and James (2003), based on distortion, a quantity that measures the average distance between each observation and its closest cluster center, is applied confirming our selection. We note, that our purpose is to identify not only the general characteristics of the synoptic conditions favoring snowfall in Athens, but also the details referring to the position, the intensity and the trajectories of the associated circulation sys- 1957-1958 1960-1961 1963-1964 1966-1967 1969-1970 1972-1973 1975-1976 1978-1979 1981-1982 1984-1985 1987-1988 1990-1991 1993-1994 1996-1997 1999- …”

Section: Methodsmentioning

confidence: 99%

The atmospheric conditions over Europe and the Mediterranean, favoring snow events in Athens, Greece

2007

View full text Add to dashboard Cite

Abstract. The 3-dimensional structure and the evolution of atmospheric circulation favoring snowfall in Athens are examined. The study refers to 61 snow events, which occurred during the period 1958-2001. For each one of the events, the patterns of MSL pressure, 850 hPa and 500 hPa air temperatures, 500 hPa geopotential height and 1000-500 hPa thickness are constructed for the European region, for the day before (D-1), the first day (D) and the day after the end of the event (END). A statistical methodology involving Factor Analysis and Cluster Analysis is applied to the above data sets and the 61 cases are finally classified into five clusters. These clusters are generally characterized by a north-easterly flow in the lower troposphere over the Athens area. This flow is associated with the presence of a low pressure system around Cyprus and an anticyclone over Europe. The position, the intensity and the trajectories of the surface and the upper air systems during D-1, D and END days are generally different among the five clusters.

show abstract

“…Moreover, after this point the curve levels off. This indicates that a further increase in the number of parameters will not significantly improve the log-likelihood (Sugar and James, 2003). We observed that, on average, the total log-likelihood of a model increases with the stability of the model.…”

Section: Families Identified Using Randomly Initialized Model (Rim)mentioning

confidence: 86%

Identifying Mycobacterium tuberculosis complex strain families using spoligotypes

Vitol

Driscoll

Kreiswirth³

et al. 2006

Infection, Genetics and Evolution

104

102

View full text Add to dashboard Cite

We present a novel approach for analysis of Mycobacterium tuberculosis complex (MTC) strain genotyping data. Our work presents a first step in an ongoing project dedicated to the development of decision support tools for tuberculosis (TB) epidemiologists exploiting both genotyping and epidemiological data. We focus on spacer oligonucleotide typing (spoligotyping), a genotyping method based on analysis of a direct repeat (DR) locus. We use mixture models to identify strain families of MTC based on their spoligotyping patterns. Our algorithm, SPOTCLUST, incorporates biological information on spoligotype evolution, without attempting to derive the full phylogeny of MTC. We applied our algorithm to 535 different spoligotype patterns identified among 7166 MTC strains isolated between 1996 and 2004 from New York State TB patients. Two models were employed and validated: a 36-component model based on global spoligotype database SpolDB3, and a randomly initialized model (RIM) containing 48 components. Our analysis both confirmed previously expert-defined families of MTC strains and suggested certain new families. SPOTCLUST, which is available online, can be further improved by incorporating data obtained using additional strain genetic markers and epidemiological information. We demonstrate on New York City (NYC) patient data how the resulting models can potentially form the basis of TB control tools using genotyping.3

show abstract

Finding the Number of Clusters in a Dataset

Cited by 735 publications

References 3 publications

A novel approach for extracting ideal exemplars by clustering for massivetime-ordered datasets

A novel approach for extracting ideal exemplars by clustering for massivetime-ordered datasets

The atmospheric conditions over Europe and the Mediterranean, favoring snow events in Athens, Greece

Identifying Mycobacterium tuberculosis complex strain families using spoligotypes

Contact Info

Product

Resources

About