| Summary: | Target selection models play an important role in business practice. They are the data-mining methods that enable firms to conduct market segmentation. Marketers apply them to customer databases to identify the profiles of consumers who are most interested in a particular offer or marketing proposition. However, both the marketing and data-mining literature indicate that there is inadequate research that compares target selection models in order to help practitioners understand how to apply them. With respect to this, the focus of this study is to provide guidance on the implementation of a collection of target selection models and to assess their comparative performance with regard to their practical usefulness. This study assesses the relative performance of the methods cluster analysis alongside multiple dicsriminant analysis (MDA), Chi-square automatic interaction detector (CHAID) and expert systems in predicting the weekly expenditure of grocery products of 9,854 consumers in the UK and develops a new approach based on fuzzy expert systems. The comparison of these methods is conducted by using three criteria (parity test, hit rate and lift charts) and one validation method (M-fold cross-validation). The results suggest that these methods vary in performance across different criteria. Overall, CHAID and fuzzy expert systems outperformed cluster analysis alongside MDA in terms of classification accuracy (parity test and the hit rate), moreover, as far as practical applicability is concerned (lift charts), no clear conclusions could be drawn between CHAID and cluster analysis alongside MDA on which of the two is best, while expert systems performed last. Furthermore, from the findings mentioned and from the empirical application of the methods examined, conclusions are derived on the features of their processes that affect their practical usefulness and on the way they should be implemented.
|
|---|
