Store Congestion Forecast under the Pandemic using Point of Sales Statistics

Under the COVID-19 pandemic, it is necessary to balance social distancing and continuous economic activities. In this study, we report on our developed service that forecasts the congestion level of regional commercial facilities using point-of-sales (POS) statistics. POS statistics data were collected for over a year from 150 commercial facilities in Tokyo. Through the analysis of a total of over 100 million customers, we clarified the factors that affect congestion levels of commercial facilities in each ward of Tokyo. Based on this analysis, we developed a congestion forecast model that predicts future congestion levels from several factors such as a big event, business restrictions, and weather. We implemented a web service incorporating this model and published estimated congestion levels both on our website and a television program. The experimental results show that the model has a high prediction accuracy with a coefficient of determination greater than 0.95 on average, which implies that big data from POS has great potential for value creation under the pandemic.