Evaluation of structure and reproducibility of cluster solutions using the bootstrap

Abstract

Segmentation results derived using cluster analysis depend on (1) the structure of the data and (2) algorithm parameters. Typically neither the data structure is assessed in advance of clustering nor is the sensitivity of the analysis to changes in algorithm parameters. We propose a benchmarking framework based on bootstrapping techniques that accounts for sample and algorithm randomness. This provides much needed guidance both to data analysts and users of clustering solutions regarding the choice of the final clusters from computations which are exploratory in nature. This is a preprint of an article that has been accepted for publication in Marketing Letters. The original publication is available at www.springerlink.com. Please use the journal version for citation, see http://dx.doi.org/10.1007/s11002-009-9083-4 for details. Disciplines Business | Social and Behavioral Sciences AbstractSegmentation results derived using cluster analysis depend on (1) the structure of the data and (2) algorithm parameters. Typically neither the data structure is assessed in advance of clustering nor is the sensitivity of the analysis to changes in algorithm parameters. We propose a benchmarking framework based on bootstrapping techniques that accounts for sample and algorithm randomness. This provides much needed guidance both to data analysts and users of clustering solutions regarding the choice of the final clusters from computations which are exploratory in nature.

Crisis-resistant tourists

“…Several cluster algorithms, including k-means, were calculated; the neural gas solution was chosen because it generated the most distinct segments. Neural gas also emerged as the most stable algorithm for this type of data in simulations on both artificial and real-world data (Dolnicar, Leisch, Weingessel, Buchta, & Dimitriadou, 1998;Dolnicar & Leisch, 2010), and has been used for market segmentation studies in tourism in the past (Mazanec, Ring, Stangl, & Teichmann, 2010).…”

“…To determine a suitable number of clusters, the bootstrapping method by Dolnicar and Leisch (2010) was used. Bootstrapping simulates what would happen if new survey data were clustered.…”

Identifying tourists with smaller environmental footprints

“…This decision was made after investigating whether or not a posteriori market segmentation would provide any additional insights. The framework proposed by Dolnicar and Leisch (2009) was used for the determination of the most suitable number of segments in a posteriori market segmentation. Achieving high reliability levels for any segment number higher than two segments would indicate that pro-environmental behavior is not sufficiently well represented by simply comparing people with a high tendency and people with a low tendency to behave in an environmentally friendly manner.…”

Market segmentation approaches in tourism

“…But in the case of most segmentation criteria, such clear natural groups rarely exist and this is increasingly acknowledged among segmentation experts (Mazanec et al 1997;Wedel and Kamakura 1998). As a consequence of this new understanding of what market segmentation can and cannot achieve, Dolnicar and Leisch (2010) introduce the terms "natural", "reproducible" and "constructive clustering / segmentation".…”

“…This procedure has been proposed and illustrated by Dolnicar and Leisch (2010) using R code which runs the repeat analysis automatically, but can be reproduced with other statistical packages. If the exact same segments emerge from repeated computations it can be assumed that natural segments exist, if similar segments emerged the segments are likely to be reproducible and if segments are different every single time, then segments need to be constructed artificially.…”