Analyzing Repurchase in Single-Item-Regular-Mail-Order with LightGBM and Logistic Regression

Abstract

In this paper, we construct a prediction model and analyze factors for repurchase in single-item-regular-mail-order (SIRMO). SIRMO is new sales model, hence there are few previous researches for analysis of SIRMO. We use LightGBM as our prediction model and it shows improvement of especially accuracy and precision for baseline. We also find characteristic factors of repurchase in SIRMO by using a logistic regression model.