A data mining framework for opitmal product selection in convenience stores

Abstract

Previous research in the field of data mining has demonstrated that the technique of association rules is very well suited to find patterns in the purchase behaviour of customers. However, practitioners occasionally criticize that it is not straightforward to adopt the discovered knowledge for concrete retail marketing decision-making. This is partially due to the difficult integration of retail domain knowledge into the mining process which sometimes causes the discovered knowledge to be sterile. This paper makes an attempt at integrating category management knowledge into the knowledge discovery process in order to obtain more useful results, i.e. results that can better be used for concrete decision-making in retailing. More specifically, an integer programming model for product selection is proposed which takes into account cross-selling effects between products and also enables the retailer to integrate category management knowledge into the model. First results on real-world retail data demonstrate the success of the approach.