ポアソン潜在ブロックモデルによるドラッグストアにおける購買データの解析

抄録

In this paper, POS data with IDs of drugstores is applied to the analysis method proposed by Govaert & Nadif (2010). By using the Poisson latent block model with snacks purchase history data at drugstores located in the Gifu region, we cluster both the customers and the snacks brands simultaneously, and grasp the snack brand group that the customer group with a high purchase frequency choose. Furthermore, based on the model estimation, we discuss effective marketing measures in drugstores. Although we analyze only snack category, it can be extended to other product categories and is an effective way to summarize Big Data.