Ranking and selection of unsupervised learning marketing segmentation

This paper addresses the problem of choosing the most appropriate classification from a given set of classifications of a set of patterns. This is a relevant topic on unsupervised systems and clustering analysis because different classifications can in general be obtained from the same data set. The provided methodology is based on five fuzzy criteria which are aggregated using an Ordered Weighted Averaging (OWA) operator. To this end, a novel multi-criteria decision making (MCDM) system is defined, which assesses the degree up to which each criterion is met by all classifications. The corresponding single evaluations are then proposed to be aggregated into a collective one by means of an OWA operator guided by a fuzzy linguistic quantifier, which is used to implement the concept of fuzzy majority in the selection process. This new methodology is applied to a real marketing case based on a business to business (B2B) environment to help marketing experts during the segmentation process. As a result, a segmentation containing three segments consisting of 35, 98 and 127 points of sale respectively is selected to be the most suitable to endorse marketing strategies of the firm. Finally, an analysis of the managerial implications of the proposed methodology solution is provided.

► Five fuzzy criteria are defined and analysed to collectively evaluate classifications.
► Fuzzy multi-criteria selection methodology for unsupervised learning systems proposed.
► Application of proposed methodology to a real B2B marketing case study.
► Analysis of managerial implications of the proposed methodology solution is provided.
► New approach avoids segmentations useful for marketing experts from being discarded.