LAUSR

Previous research on RFM (recency, frequency, and monetary value) models focused on only one type of user behavior data, i.e., the purchase behavior, without considering the interactions between users and items, such as clicking, favorite, and adding to cart. In this study, we propose a novel solution for deconstructing the multiple behaviors of consumers in a specific period and performing customer segmentation in an application promotion system called multi-behavior RFM (MB-RFM) based on the self-organizing map (SOM) algorithm. First, using the R, F, and M values, we analyzed the weight relationship between multiple behaviors of users and items using the superiority chart and entropy value methods. Each behavior ascribed to a customer was considered to be a part of MB-RFM model values, which were then used to classify customers using an improved SOM neural network. Subsequently, various promotion strategies were developed according to customer categories that can help application vendors in improving their application utilization and implementing targeted promotion strategies. To prove the effectiveness of the proposed method in sparse datasets, two real-world datasets were used to perform experiments, whose results demonstrated that the classification performance of our method was significantly more accurate.