潜在クラスモデルを利用した取引データのセグメンテーション

抄録

In this article, we proposed a segmentation method of transaction data by making use of a Latent Class Model. In the analysis, not only a standard Latent Class Model, but also a model that relaxed the assumption of local independence was applied to the transaction data of a convenience store.
Consequently, based on the AIC criterion, a model with six classes was adopted. All of these segments were interpreted as having a different purchace-intentions. Additionally, the purchase-intention of a coming-into-the-store visitor and the size of the purchase could be guessed from transaction data. Furthermore, it was shown that the information acquired by analyzing transaction data by our method may translate into useful information for working on a store policy corresponding to the characteristics of the coming-to-the-store visitors of a convenience store from a practical viewpoint.