F-RFM-Miner: an efficient algorithm for mining fuzzy patterns using the recency-frequency-monetary model

Yanlin Qi, Fuyin Lai, Guoting Chen, Wensheng Gan

Abstract: In database marketing, recency, frequency, and monetary (RFM) analysis is an important tool to segment customers based on their recent purchase behaviors. By combining the RFM model with frequent pattern mining algorithms such as RFMP-Growth and fuzzy-RFU-tree, RFM-patterns can be mined. RFMP-Growth and fuzzy-RFU-tree use a tree-based structure; however, patterns found by RFMP-Growth do not contain qualitative information among items. By applying the fuzzification method, patterns mined by fuzzy-RFU-tree contain qualitative information about items. However, this algorithm consumes considerable memory and time. Therefore, to discover valuable fuzzy-RFM-patterns efficiently, we first introduce a list structure and propose the F-RFM-Miner algorithm. Consequently, we design two new pruning strategies to reduce the number of candidate patterns. Moreover, we conduct experiments on dense and sparse datasets to compare our algorithm with state-of-the-art algorithm and test the efficiency of the new pruning strategies. The experiment results show that F-RFM-Miner performs better than fuzzy-RFU-tree.