<p>Effective planning of public charging stations (PCSs) is crucial for accelerating the transition to an electrified transportation system. However, existing approaches often overlook the behavioral complexity of battery electric vehicle (BEV) users, leading to inefficient and inequitable infrastructure allocation. This study proposes a user-centered analytical framework to examine the supply–demand dynamics and spatial equity of PCS deployment. Using trajectory and charging data from over 150,000 operational BEVs in Shanghai, a two-dimensional Latent Dirichlet Allocation (LDA) model identifies distinct user subgroups with unique spatiotemporal charging patterns. For each subgroup, an interpretable machine learning model (Random Forest) is developed to uncover the key factors influencing PCS choice, yielding higher predictive accuracy than models trained on the aggregated dataset. The framework further estimates potential charging demand under an equitable PCS capacity scenario to detect supply–demand imbalances. The findings provide actionable insights for optimizing PCS deployment, improving resource utilization, and promoting equitable access to charging infrastructure in rapidly electrifying cities.</p>