The Effects of Initial Values and the Covariance Structure on the Recovery of some Clustering Methods

Hajnal, István; Loosveldt, Geert

doi:10.1007/978-3-642-59789-3_7

Cited by 2 publications

(3 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Other limitations of the K‐means method include sensitivity to initial values and outliers 33 ; local optima can be generated for certain initial values. In our study, we initialized the centers with multiple random seeds 40,41 and obtained stable clustering results. For dealing with outliers, we excluded records with mean dose greater than 5.0 DDD during any

τ

‐day episode.…”

Section: Discussionmentioning

confidence: 99%

Classification of drug use patterns

Righolt

Zhang

Mahmud

2020

Pharmacology Res & Perspec

View full text Add to dashboard Cite

Characterizing long‐term prescription data is challenging due to the time‐varying nature of drug use. Conventional approaches summarize time‐varying data into categorical variables based on simple measures, such as cumulative dose, while ignoring patterns of use. The loss of information can lead to misclassification and biased estimates of the exposure‐outcome association. We introduce a classification method to characterize longitudinal prescription data with an unsupervised machine learning algorithm. We used administrative databases covering virtually all 1.3 million residents of Manitoba and explicitly designed features to describe the average dose, proportion of days covered (PDC), dose change, and dose variability, and clustered the resulting feature space using K‐means clustering. We applied this method to metformin use in diabetes patients. We identified 27,786 metformin users and showed that the feature distributions of their metformin use are stable for varying the lengths of follow‐up and that these distributions have clear interpretations. We found six distinct metformin user groups: patients with intermittent use, decreasing dose, increasing dose, high dose, and two medium dose groups (one with stable dose and one with highly variable use). Patients in the varying and decreasing dose groups had a higher chance of progression of diabetes than other patients. The method presented in this paper allows for characterization of drug use into distinct and clinically relevant groups in a way that cannot be obtained from merely classifying use by quantiles of overall use.

show abstract

τ

‐day episode.…”

Section: Discussionmentioning

confidence: 99%

Classification of drug use patterns

Righolt

Zhang

Mahmud

2020

Pharmacology Res & Perspec

View full text Add to dashboard Cite

show abstract

“…Cerioli and Zani (2001) proposed a density search method based on quadrant counts arising from multivariate histograms to (a) help determine the number of clusters and (b) restrict initial centroids to be chosen from areas of high density; Faber (1994) suggested choosing the starting points uniformly randomly to give preference to points in dense regions. Hajnal and Loosveldt (2000) suggested that the initialization method implemented by SAS is superior to using a random seed as starting value for K-means. However, in a much broader study examining the performance of the default functions for several commercial software packages (i.e.…”

Section: Methods Of Initializationmentioning

confidence: 99%

“…Hajnal and Loosveldt (2000) suggested that the initialization method implemented by SAS is superior to using a random seed as starting value for K ‐means. However, in a much broader study examining the performance of the default functions for several commercial software packages (i.e.…”

Section: Important Considerationsmentioning

confidence: 99%

K‐means clustering: A half‐century synthesis

Steinley

2006

Brit J Math & Statis

819

550

View full text Add to dashboard Cite

This paper synthesizes the results, methodology, and research conducted concerning the K-means clustering method over the last fifty years. The K-means method is first introduced, various formulations of the minimum variance loss function and alternative loss functions within the same class are outlined, and different methods of choosing the number of clusters and initialization, variable preprocessing, and data reduction schemes are discussed. Theoretic statistical results are provided and various extensions of K-means using different metrics or modifications of the original algorithm are given, leading to a unifying treatment of K-means and some of its extensions. Finally, several future studies are outlined that could enhance the understanding of numerous subtleties affecting the performance of the K-means method.

show abstract

The Effects of Initial Values and the Covariance Structure on the Recovery of some Clustering Methods

Cited by 2 publications

References 8 publications

Classification of drug use patterns

Classification of drug use patterns

K‐means clustering: A half‐century synthesis

Contact Info

Product

Resources

About