The use of machine learning to predict the main factors that influence the continuous usage of mobile food delivery apps

Abstract

The popularity of mobile food delivery apps (MFDAs) and the online food delivery industry surged during the COVID-19 epidemic. Despite the explosive growth in the use of these apps, relatively limited research has been done to determine what affects their continuous use. This study predicts the continuous use of MFDAs and explores the variables that influence this utilization using a novel machine learning (ML) based approach. The machine learning models included four distinct constructs (i.e., features): perceived compatibility, convenience, online reviews, and delivery experience. These features were measured using a survey instrument. Eight different machine learning (ML) models, ranging from basic decision trees to neural networks, were deployed. All eight models achieved high prediction accuracy of above 93%, with the CatBoost model having the highest accuracy among them at 98%. Feature importance analysis revealed perceived compatibility to be the most important factor impacting the continuous usage of MFDAs followed by convenience, online reviews, and delivery experience respectively. The study’s findings have ramifications for MFDA marketing and design. Given the significance of perceived compatibility, MFDA marketing campaigns should have a strong emphasis on highlighting how well these apps fit with the users’ lifestyles.