This study investigates the multifaceted nature of poverty in the United States by applying machine learning models based on Support Vector Machines (SVMs) to a large, cross-sectional dataset drawn from the IPUMS database. The primary aim is to identify and rank the significance of social, demographic, economic, and geographic factors influencing poverty. Three SVM models were trained using categorized variables—predisposing, socio-demographic, and socio-economic—and evaluated using cross-validation and performance metrics such as accuracy, AUC, and F1 score. Sensitivity and Variable Effect Characteristic (VEC) analyses further highlighted the dominant influence of total personal income, employment status, and educational attainment on poverty status. Additionally, racial, and geographic factors also played substantial roles, underlining systemic disparities. The findings reinforce the importance of using complex, multi-variable models to understand poverty and support the formulation of more comprehensive, equity-focused policy interventions.