Predicting Grocery Store Visits During the Early Outbreak of COVID-19 with Machine Learning

Area of Science:

• Epidemiology
• Mobility Studies
• Data Science

Background:

• The COVID-19 pandemic necessitated changes in essential activities like grocery shopping.
• Understanding shifts in grocery store access is crucial during public health crises.

Purpose of the Study:

• To analyze changes in grocery store visits during the early COVID-19 pandemic.
• To develop a predictive model for future grocery shopping behavior within the pandemic's initial phase.

Main Methods:

• Analysis of grocery store visit data (in-store and pickup) from February 15 to May 31, 2020, across six US states/counties.
• Utilized Google Mobility Reports and a long short-term memory (LSTM) network for predictive modeling.

Main Results:

• Grocery visits initially increased over 20% post-national emergency declaration, then fell below baseline within a week.
• Weekend shopping was more affected than weekday shopping early in the pandemic.
• Some states showed recovery by May, but certain urban areas lagged.

Conclusions:

• The study provides insights into mobility patterns related to grocery shopping during the pandemic.
• The LSTM model demonstrated effectiveness in predicting general trends in grocery visit changes.