Exemplar-Based Clustering via Simulated Annealing

Brusco, Michael J.; Köhn, Hans-Friedrich

doi:10.1007/s11336-009-9115-2

Cited by 20 publications

(29 citation statements)

References 57 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Brusco and Köhn () showed that the affinity propagation algorithm often failed to obtain a globally optimal solution even for relatively small problem instances from the literature. Similar findings have been reported by Brusco and Köhn () and Brusco and Steinley (). Nevertheless, it should be acknowledged that suboptimality issues have commonly plagued traditional methods such as Ward's method and K ‐means clustering (Brusco & Cradit, ; Brusco & Steinley, ; Steinley, , ), yet such methods have been gainfully employed for several decades.…”

Section: Merits Of Affinity Propagation For Psychological Researchsupporting

confidence: 92%

“…Brusco and Köhn () showed that the number of clusters could vary drastically (from just a few clusters to several hundred) as the preference values were changed. More recently, Brusco and Steinley () conducted a simulation‐based evaluation of different multiples of the median for setting the preference vector.…”

Section: Underlying Optimization Modelmentioning

confidence: 99%

“…At the same time, some authors have explored alternatives for specifying p. For example, Wang, Qin, and Zhang (2010) used a particle swarm method to search for an optimal set of preferences. Brusco and K€ ohn (2009) showed that the number of clusters could vary drastically (from just a few clusters to several hundred) as the preference values were changed. More recently, Brusco and Steinley (2015) conducted a simulation-based evaluation of different multiples of the median for setting the preference vector.…”

Section: Preferences and Similarity Data Issuesmentioning

confidence: 99%

See 2 more Smart Citations

Affinity propagation: An exemplar‐based tool for clustering in psychological research

Brusco

Steinley

Stevens

et al. 2018

Brit J Math & Statis

Self Cite

View full text Add to dashboard Cite

Affinity propagation is a message-passing-based clustering procedure that has received widespread attention in domains such as biological science, physics, and computer science. However, its implementation in psychology and related areas of social science is comparatively scant. In this paper, we describe the basic principles of affinity propagation, its relationship to other clustering problems, and the types of data for which it can be used for cluster analysis. More importantly, we identify the strengths and weaknesses of affinity propagation as a clustering tool in general and highlight potential opportunities for its use in psychological research. Numerical examples are provided to illustrate the method.

show abstract

Section: Merits Of Affinity Propagation For Psychological Researchsupporting

confidence: 92%

Section: Underlying Optimization Modelmentioning

confidence: 99%

Section: Preferences and Similarity Data Issuesmentioning

confidence: 99%

See 1 more Smart Citation

Affinity propagation: An exemplar‐based tool for clustering in psychological research

Brusco

Steinley

Stevens

et al. 2018

Brit J Math & Statis

Self Cite

View full text Add to dashboard Cite

show abstract

“…Therefore, multiple restarts of the algorithm are advised. Hansen and Mladenović (1997) indicated that the multiple restart approach was generally effective for problems where K < 50 (see also Brusco & Köhn. 2009).…”

Section: Methodsmentioning

confidence: 99%

A comparison of latent class, K-means, and K-median methods for clustering dichotomous data.

Brusco

Shireman

Steinley

2017

Psychological Methods

Self Cite

View full text Add to dashboard Cite

The problem of partitioning a collection of objects based on their measurements on a set of dichotomous variables is a well-established problem in psychological research, with applications including clinical diagnosis, educational testing, cognitive categorization, and choice analysis. Latent class analysis and K-means clustering are popular methods for partitioning objects based on dichotomous measures in the psychological literature. The K-median clustering method has recently been touted as a potentially useful tool for psychological data and might be preferable to its close neighbor, K-means, when the variable measures are dichotomous. We conducted simulation-based comparisons of the latent class, K-means, and K-median approaches for partitioning dichotomous data. Although all three methods proved capable of recovering cluster structure, K-median clustering yielded the best average performance, followed closely by latent class analysis. We also report results for the three methods within the context of an application to transitive reasoning data, where it was found that the three approaches can exhibit profound differences when applied to real data.

show abstract

“…To select an optimal number of clusters, we performed visual analysis of a dendrogram representing the structure of the data and confirmed that we could identify natural groupings using bivariate matrices which provided construct validation. Next, we used the k-medians cluster methodology with a Euclidean distance similarity measure (L2) to verify cluster classifications, for a set number of k clusters ranging from 3–8 (Brusco & Köhn, 2009; Kohn et al, 2010). We used the Calinski pseudo-F statistic, which measures the ratio of between cluster variance to within cluster variance as a quantitative measure of the distinctness of the groups generated by the cluster analysis and provide a stopping rule to optimize the number of groups selected (Calinski & Harabasz, 1974).…”

Section: Methodsmentioning

confidence: 99%

Residential patterns in older homeless adults: Results of a cluster analysis

Lee

Guzman

Ponath

et al. 2016

Social Science & Medicine

View full text Add to dashboard Cite

Adults aged 50 and older make up half of individuals experiencing homelessness and have high rates of morbidity and mortality. They may have different life trajectories and reside in different environments than do younger homeless adults. Although the environmental risks associated with homelessness are substantial, the environments in which older homeless individuals live have not been well characterized. We classified living environments and identified associated factors in a sample of older homeless adults. From July 2013 to June 2014, we recruited a community-based sample of 350 homeless men and women aged fifty and older in Oakland, California. We administered structured interviews including assessments of health, history of homelessness, social support, and life course. Participants used a recall procedure to describe where they stayed in the prior six months. We performed cluster analysis to classify residential venues and used multinomial logistic regression to identify individual factors prior to the onset of homelessness as well as the duration of unstable housing associated with living in them. We generated four residential groups describing those who were unsheltered (n=162), cohabited unstably with friends and family (n=57), resided in multiple institutional settings (shelters, jails, transitional housing) (n=88), or lived primarily in rental housing (recently homeless) (n=43). Compared to those who were unsheltered, having social support when last stably housed was significantly associated with cohabiting and institution use. Cohabiters and renters were significantly more likely to be women and have experienced a shorter duration of homelessness. Cohabiters were significantly more likely than unsheltered participants to have experienced abuse prior to losing stable housing. Pre-homeless social support appears to protect against street homelessness while low levels of social support may increase the risk for becoming homeless immediately after losing rental housing. Our findings may enable targeted interventions for those with different manifestations of homelessness.

show abstract

Exemplar-Based Clustering via Simulated Annealing

Abstract: cluster analysis, partitioning, heuristics, p-median model, simulated annealing,

Cited by 20 publications

References 57 publications

Affinity propagation: An exemplar‐based tool for clustering in psychological research

Affinity propagation: An exemplar‐based tool for clustering in psychological research

A comparison of latent class, K-means, and K-median methods for clustering dichotomous data.

Residential patterns in older homeless adults: Results of a cluster analysis

Contact Info

Product

Resources

About