Remaining within-cluster heterogeneity: a meta-analysis of the "dark side" of clustering methods

Abstract

In a meta-analysis of articles employing clustering methods, we find that little attention is paid to remaining within-cluster heterogeneity and that average values are relatively high. We suggest addressing this potentially problematic "dark side" of cluster analysis by providing two coefficients as standard information in any cluster analysis findings: a goodness-of-fit measure and a measure which relates explained variation of analysed empirical data to explained variation of simulated random data. The second coefficient is referred to as the Index of Clustering Appropriateness (ICA). Finally, we develop a classification scheme depicting acceptable levels of remaining within-cluster heterogeneity.

Perspective: Understanding the Nature and Measurement of the Lead User Construct

“…High expected benefit from innovative solutions means that there is high problem-solving pressure on the part of the user. As users in a market-and even within market segments-tend to exhibit high heterogeneity with respect to their needs (Franke, Reisinger, and Hoppe, 2009;Franke and von Hippel, 2003), standard products leave behind many users whose needs are not adequately addressed. Such users would benefit greatly from a tailored solution.…”

Bagged fuzzy clustering for fuzzy data: An application to a tourism market

“…The items were measured on a bipolar 7-point Likert scale anchored on [1] Offers very little and [7] Offers very much. Socio-demographics, travelling characteristics, and information sources most likely to use to plan a trip to Western Europe were also measured.…”

“…Cluster analysis has been criticized for its overestimation of the validity of the segmentation results [2] and the resulting clusters have been termed "convenient fictions" [6], a marketing term that refers to the fact that no "natural groupings" could exist, and some information is inevitably lost when objects are grouped. Information loss is not problematic per se, but it can result in the wrong conclusions [7]. Every clustering algorithm has advantages and drawbacks and has to be chosen with awareness of its characteristics and limitations [1,2].…”

“…In overlapping algorithms, an object may belong to more than one cluster [4]. Similar to overlapping algorithms, fuzzy algorithms allow objects to belong to more than one cluster and, in addition, assign to each object a degree of membership to each segment [1,7].…”

Segmenting visitors of cultural events: The case of Christmas Market

“…Among a posteriori segmentation approaches, cluster analysis remains the most popular method and the most frequently used in the literature (Jain, 2010;Dolnicar, 2002;Wedel & Kamakura, 2000). Clustering methods are generally divided into three categories: non-overlapping algorithms (each object is part of a single segment - Tuma et al, 2011); overlapping algorithms (an object may belong to more than one cluster -Wedel & Kamakura, 2000); fuzzy algorithms (each object is assigned by a degree of membership to a segment - Franke et al, 2009;Tuma et al, 2011). Hierarchical (agglomerative) and non-hierarchical (partitioning) methods are two common approaches that can be classified within non-overlapping algorithms.…”