The COVID-19 pandemic has significantly changed consumer behavior, leading to a sharp rise in the use of online home delivery services. These changes have important implications for transportation systems, logistics, and urban planning. This study aims to identify key sociodemographic and behavioral factors associated with increased delivery use during the pandemic by developing and comparing several machine learning classification models. Using data from the 2022 National Household Travel Survey (NHTS), we applied a structured preprocessing pipeline, including cleaning, merging household and person-level data, and assessing multicollinearity. Five models were trained and evaluated: Logistic Regression, Naïve Bayes, Random Forest, XGBoost, and LightGBM. LightGBM achieved the highest accuracy (97%), followed closely by XGBoost and Random Forest. Feature importance analysis based on LightGBM revealed that age, household income, education level, and prior use of delivery services were the most influential predictors of behavioral change. Descriptive analysis confirmed that younger, higher-income, and more educated individuals were more likely to increase their use of food and grocery delivery services. The findings highlight the usefulness of machine learning in identifying complex, nonlinear relationships in large-scale behavioral datasets. This study contributes evidence that sociodemographic and behavioral patterns played a crucial role in online delivery adoption during the pandemic. These insights can support transportation planners, policymakers, and logistics providers in designing adaptive systems that respond to evolving consumer needs in both emergency and post-pandemic settings.