Fast RFM Model for Customer Segmentation

Abstract

With booming e-commerce and World Wide Web (WWW), a powerful tool in customer relationship management (CRM), called the RFM analysis model, has been used to ensure that major enterprises make more profit. Combined with data mining technologies, the CRM system can automatically predict the future behavior of customers to raise customer retention rate. However, a key issue is that the existing RFM analysis models are not efficient enough. Thus, in this study, a fast algorithm based on a compact list-based data structure is proposed along with several efficient pruning strategies to address this issue. The new algorithm considers recency (R), frequency (F), and monetary/utility (M) as three different thresholds to discover interesting patterns where the R, F, and M thresholds combined are no less than the user-specified minimum values. More significantly, the downward-closure property of frequency and monetary metrics are utilized to discover super-itemsets. Then, an extensive experimental study demonstrated that the algorithm outperforms state-of-the-art algorithms on various datasets. It is also demonstrated that the proposed algorithm performs well when considering the frequency metric alone.