As natural disasters intensify, translating awareness into concrete preparedness remains difficult. We propose a framework to estimate individuals’ perceived burdens for disaster preparedness by integrating location data with psychological indicators. A BERT-based location encoder models GPS trajectories as token sequences to learn contextual mobility embeddings, which are combined with daily mobility, demographics, personality, and preparedness stage. Using five months of GPS data from 61,846 users for pre-training and survey data from 500 participants for classification, the framework estimates perceived burdens across eight behaviors. Three main insights emerged. First, mobility-informed modeling effectively estimated preparedness-related attitudes at the individual level, with higher accuracy for routine behaviors such as checking home safety or stocking supplies than for socially coordinated actions like community drills. Second, incorporating temporal and personal contextual embeddings—particularly home and workplace locations—yielded the best performance, suggesting that personal spatial context captures lifestyle patterns relevant to perceived burden. Third, adding psychological and attitudinal features, including GRIT, Big Five traits, and preparedness stage, improved classification accuracy beyond mobility features alone. Overall, these findings indicate that combining mobility and psychological factors supports more precise, personalized modeling of preparedness tendencies, providing a foundation for adaptive persuasive systems that encourage proactive, user-centered disaster preparedness.