A New Variable Weighting and Selection Procedure for <i>K</i>-means Cluster Analysis

Steinley, Douglas; Brusco, Michael J.

doi:10.1080/00273170701836695

Cited by 70 publications

(77 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…It is therefore worthwhile carefully selecting variables to be included in the segmentation base, rather than including an entire question battery by default. Noisy variables in the segmentation base can be avoided by (1) identifying and removing them after data collection (Brusco and Cradit 2001;Carmone, Kara, and Maxwell 1999;Steinley and Brusco 2008) or by (2) ensuring, before data collection, that survey questions are only included if they contain relevant information, as advocated by Rossiter (2002Rossiter ( , 2011. Methods for identifying and removing can either be employed before the clustering using characteristics of the distribution of the single variables (Steinley and Brusco 2008) or simultaneously during clustering, by taking into account the concordance and agreement between cluster solutions implied by different variables (see Brusco and Cradit 2001, who directly build on and improve Carmone, Kara, and Maxwell 1999).…”

Section: Discussionmentioning

confidence: 99%

Required Sample Sizes for Data-Driven Market Segmentation Analyses in Tourism

Dolničar

Grün

Leisch

et al. 2013

Journal of Travel Research

183

106

View full text Add to dashboard Cite

Data analysts in industry and academia make heavy use of market segmentation analysis to develop tourism knowledge and select commercially attractive target segments. Within academic research alone, approximately 5% of published articles use market segmentation. However, the validity of data-driven market segmentation analyses depends on having available a sample of adequate size. Moreover, no guidance exists for determining what an adequate sample size is. In the present simulation study using artificial data of known structure, the impact of the difficulty of the segmentation task on the required sample size is analyzed in dependence of the number of variables in the segmentation base. Under all simulated data circumstances, a sample size of 70 times the number of variables proves to be adequate. This finding is of substantial practical importance because it will provide guidance to data analysts in academia and industry who wish to conduct reliable and valid segmentation studies.

show abstract

Section: Discussionmentioning

confidence: 99%

Required Sample Sizes for Data-Driven Market Segmentation Analyses in Tourism

Dolničar

Grün

Leisch

et al. 2013

Journal of Travel Research

183

106

View full text Add to dashboard Cite

show abstract

“…For this any clustering algorithm on multivariate data can be used. Since the time complexity of this step depends on the number of observation N and of variables P we decide to screen out irrelevant features using a feature selection algorithm for unsupervised learning described in [33]. Besides the reduction of the computation time, feature selection allows also to gain in interpretability of the clustering since it highly depends on the data.…”

Section: Clustering Electrical Load Curvesmentioning

confidence: 99%

Scalable Clustering of Individual Electrical Curves for Profiling and Bottom-up Forecasting

Auder¹,

Cugliari²,

Goude³

et al. 2018

Preprint

View full text Add to dashboard Cite

Smart grids require flexible data driven forecasting methods. We propose clustering tools for bottom-up short-term load forecasting. We focus on individual consumption data analysis which plays a major role for energy management and electricity load forecasting. The two first sections are dedicated to the industrial context and a review of individual electrical data analysis. We are interested in hierarchical time-series for bottom-up forecasting. The idea is to disaggregate the signal in such a way that the sum of disaggregated forecasts improves the direct prediction. The 3-steps strategy defines numerous super-consumers by curve clustering, builds a hierarchy of partitions and selects the best one minimizing a forecast criterion. Using a nonparametric model to handle forecasting, and wavelets to define various notions of similarity between load curves, this disaggregation strategy applied to French individual consumers leads to a gain of 16% in forecast accuracy. We then explore the upscaling capacity of this strategy facing massive data and implement proposals using R, the free software environment for statistical computing. The proposed solutions to make the algorithm scalable combines data storage, parallel computing and double clustering step to define the super-consumers.Keywords: Clustering; Forecasting; Hierarchical Time-Series; Individual Electrical Consumers; Scalable; Short Term; Smart Meters; Wavelets Industrial contextEnergy systems are facing a revolution and many challenges. On the one hand, electricity production is moving to more intermittency and complexity with the increase of renewable energy and the development of small distributed production units such as photovoltaic panels or wind farms. On the other hand, consumption is also changing with plug-in (hybrid) electric vehicles, heat pumps, the development of new technologies such as smart phones, computers, robots that often come with batteries. To maintain the electricity quality, energy stakeholders are developing smart grids (see [1,2] This results into new opportunities such as local optimisation of the grid, demand side management and smart control of storage devices. Exploiting the smart grid efficiently requires advanced data analytics and optimisation techniques to improve forecasting, unit commitment, and load planning at different geographical scales. Massive data sets are and will be produced as Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted:

show abstract

“…Segundo Steinley & Brusco (2008a) e Anzanello & Fogliatto (2011), uma maior variância sugere variáveis mais dispersas e, em consequência, com maior capacidade de diferenciarem observações em grupos quando comparadas a variáveis com menores variâncias. O número de componentes a ser retido pode ser definido através do Scree Graph ou de validação cruzada (Duda et al, 2001).…”

Section: Passo 2 -Aplicação Da Acp Nos Dados Remapeados E Geração Do unclassified

Seleção de variáveis para clusterização de bateladas produtivas através de ACP e remapeamento kernel

Cervo

Anzanello

2015

Prod.

View full text Add to dashboard Cite

ResumoTécnicas de clusterização visam à formação de grupos de observações homogêneas dentro de um mesmo grupo e significativamente distintas das observações inseridas em outros grupos. Em processos industriais cuja produção é apoiada em bateladas, a definição de famílias (grupos) de bateladas com perfis semelhantes auxilia na definição de estratégias de controle e monitoramento desses processos. Este artigo propõe um método para seleção das variáveis de clusterização mais relevantes para formação de famílias de bateladas. Para tanto, integra funções kernel a um novo índice de importância de variáveis gerado a partir dos parâmetros oriundos da Análise de Componentes Principais (ACP). A qualidade dos agrupamentos formados é avaliada através do Silhouette Index (SI). Quando aplicada em três processos produtivos, a sistemática proposta reteve em média 5,16% das variáveis iniciais e elevou o SI médio em 235,4% frente à utilização de todas as variáveis. Um estudo de simulação também é realizado para avaliar a robustez do método. Palavras-chaveAnálise de clusterização. Seleção de variáveis. Kernel. Processos em batelada.

show abstract

A New Variable Weighting and Selection Procedure for K-means Cluster Analysis

Cited by 70 publications

References 31 publications

Required Sample Sizes for Data-Driven Market Segmentation Analyses in Tourism

Required Sample Sizes for Data-Driven Market Segmentation Analyses in Tourism

Scalable Clustering of Individual Electrical Curves for Profiling and Bottom-up Forecasting

Seleção de variáveis para clusterização de bateladas produtivas através de ACP e remapeamento kernel

Contact Info

Product

Resources

About