Abstract: Strategic early-warning is crucial for ensuring national security and regional stability. Addressing the issues of strong subjectivity and high costs in existing early-warning modes, a strategic symptom early-warning framework that integrates expert knowledge with large language models（LLMs）is designed. Open-source intelligence texts are regarded as data support and the Israeli-Palestinian conflict is taken as the strategic background. Firstly, expert knowledge is utilized to establish an event ontology model, and to define event types, and to construct an event causal graph. Subsequently, an LLMs serves as the foundation toaccomplish event abstracts, classification, extraction, and matching. Finally, reasoning tools are employed by LLMs to predict the occurring probability of symptoms and provide interpretations of the warning results. The case analysis demonstrates that the proposed framework can generate strategic early warnings and explanations for assistant decision-making and can reflect the changes in the strategic situation.