Demand Forecasting for Canned Beer Using Retail Store POS Data and SNS Data

Abstract

In recent years, the proliferation of social networks and news media has brought about changes in consumer behavior at an unprecedented speed due to societal circumstances, trends, and seasonality. Particularly, the demand for inexpensive luxury goods and daily necessities can be greatly influenced by the social environment, making it crucial for retailers and manufacturers to quickly identify signs of change. This study constructs a demand forecasting model for canned beer, a representative luxury good, using POS data from multiple large retail stores and SNS data. The proposed method suggests a way to extract features that reflect changes in the social environment from SNS post texts. The results of the experiments confirmed that the features obtained by the proposed method improved the predictive performance of models that use only sales data as input.