Using AI Models to Understand the Impact of COVID-19 in the Context of Long COVID and Food Delivery Operations

Abstract

Long COVID is unlike the usual diseases with well-defined symptoms or any medical test parameters. According to the CDC (Centers for Disease Control and Prevention), Long COVID is broadly defined as signs, symptoms, and conditions that continue or develop after acute COVID-19 infection. The definition is so broad that any two Long COVID patients may have entirely dissimilar symptoms or medical problems and this makes it almost impossible to come up with any discernible means to predict Long COVID. This paper proposes a novel method to predict Long COVID by using network analytics and machine learning. Our methodology has no medical basis but the methodology used can help in furthering the research by medical experts. The usual method of dimensionality reduction such as PCA (Principal Component Analysis) did not help much. Community Detection algorithms using a bi-partite network between COVID-19 and Long COVID patients provide extra information to remove features from over 3,500 to fewer than 300. The dimensionality reduction obtained in such a manner coupled with improvement in the ratio of Long COVID to Non-Long COVID cases in the datasets by controlling the number of Non-Long COVID cases has a strong impact on the performance of machine learning models. We identified two demographic groups for our study adult and pediatric because of differences in their vulnerabilities and immunity levels. Although the results of network analytics are different, prediction accuracies through LSTM and neural networks are above 90% in both cases.

COVID-19 has also had an adverse global impact on various industry sectors. It has led to significant changes in societal behavior. Social distancing made public spaces hazardous, shifting consumer habits, including purchasing and spending patterns. People have been driven toward online resources and delivery services, causing disruptions and impacting industries. The study investigates and identifies new online food delivery patterns that emerged during COVID-19. We focus on the food delivery industry in a University town, integrating 183 restaurants to understand how e-commerce and consumer behavior with respect to restaurant food delivery changed from pre-COVID to the COVID-19 times. We use AI and machine learning techniques to collect and analyze data collected over three years. Findings suggest that new emerging patterns require adaption to the variability in the types of food that consumers are ordering, ordering times, delivery locations, etc. Such insights provide for resource planning and allocation decisions.