Financial literacy is essential for sound financial decision-making and individual well-being. Identifying individuals with low financial literacy is crucial for designing targeted educational programs. Recent advances in statistics and machine learning provide powerful tools for predicting financial literacy levels based on demographic, education and other socioeconomic factors. This study applies supervised learning algorithms to predict financial knowledge (FK) using survey data on young adults in Italy, provided by the Bank of Italy. To enhance model interpretability, we employ SHAP (SHapley Additive exPlanations) analysis, which identifies key drivers of FK predictions. Our results suggest that younger adults, particularly women and those from lower socioeconomic backgrounds, are at higher risk of low FK scores. This is often linked to gaps in numeracy and digital skills, as well as to lower responsibilities in daily financial decision-making. These findings highlight the importance of targeted educational initiatives that aim to address these deficiencies and improve FK among vulnerable populations.