Unsupervised Learning and Market Basket Analysis in Market Segmentation
Lecture Notes in Engineering and Computer Science
Identifying and classifying market segments often involves a 'best-guess' for marketing and sales managers. This leads to a set of assumptions regarding the type and importance of customer variables and associated data. The age of 'big data' now provides an opportunity to start with the data and to then work backwards - i.e. to 'let the data tell you'. While these techniques have the advantages of reducing user bias, they are still at their infancy. In this study, we develop a two-stage approach using a large set of point-of- sale (POS) data to i) segment a retail market and ii) identify relative segment purchase probabilities. Stage one involves an unsupervised learning approach based on three purchase characteristics (Recency, Frequency and Monetary value - RFM) and product attributes to identify segments in the customer dataset. Stage two involves Market Basket Analysis (MBA) to determine the likely probabilities of purchase behaviors for each segment. Given these outcomes, we argue that marketing and sales managers have a more robust method for identifying market segments and associated purchase behaviors through predictive analytics.
Open Access Status
This publication is not available as open access