Purchase Product Classification by STL Decomposition of Time-series Sales Data

Abstract

Product sales in supermarkets are affected by a variety of external factors. One of the key insights in marketing is to infer the impact of these external factors from sales information and to determine the sensitivity to various factors. In this study, based on food sales data in supermarkets over a five-year period from 2018, we attempted to classify products by the degree of influence of external factors such as COVID-19 by using STL decomposition to isolate seasonal variations and residual components and clustering these components. The results shows that there are some products that were significantly affected by external factors. These product-specific characteristics can be expected to provide useful information for future marketing and sales strategies.