In-store shopping trip predictions and impact factors during COVID-19 emergencies

ABSTRACT

Despite the rapid growth of online shopping during COVID-19, a significant number of consumers still prefer in-store shopping. This study leverages two years (i.e. pre-pandemic and pandemic) of smartphone location data to develop machine learning (ML) models, specifically Random Forest (RF) and Extreme Gradient Boosting (XGBoost), for predicting community (e.g. block group (BG)) level in-store shopping trips for department stores, shopping malls, supermarkets, and wholesale stores. This study identifies that temperature, accessibility to stores, and the number of online shopping last-mile delivery are the three most important factors influencing shopping trips; specifically, the extent of online shopping is a critical determinant for supermarkets and wholesale store trip-makings before and during the pandemic. The models developed and important determinants of shopping trips will provide useful insight for shopping trip demand forecasting as well as impact assessments of relevant policies on in-store shopping demand during emergencies.

KEYWORDS:

• COVID-19
• shopping trips
• e-commerce
• machine learning
• shopping facilities

Acknowledgements

This project was funded by the TranSET (21ITSUTA03), a U.S. DOT University Transportation Center.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Tran-SET – Transportation Consortium of South-Central States [grant number 21ITSUTA03].