Sequential Recommendation, Biomedical Risk Evidence, and Pre-Disaster Relocation Modeling

Abstract

Sequential recommendation and disaster relocation modeling both address dynamic decision processes under uncertainty. LLM-based recommendation systems use user histories and dynamic indexing mechanisms to improve prediction of future preferences, while pre-disaster relocation models represent household decisions before flood events. Both domains require attention to temporal behavior, heterogeneous agents, and context-sensitive decision pathways. Biomedical evidence on hepatic ischemia-reperfusion and remote cardiac injury adds a mechanistic risk dimension, illustrating how complex biological processes can inform intelligent decision systems. Universal transferable adversarial attacks further highlight the need to protect such systems from manipulation when they are deployed in high-stakes environments. This literature cluster supports a broad view of intelligent decision modeling that integrates sequential behavior, biomedical risk evidence, disaster adaptation, and adversarial robustness.