Abstract: To address the problems that the compilation of operation and maintenance (O&M) plans for railway information systems mainly relied on manual experience and static regulations, suffered from fixed formats and contents, and hardly adapted to practical O&M work, this paper proposed an automatic compilation method for railway information system O&M plans integrating large language models and association rules. The study used a large language model to generate outline O&M plans. On the basis of the outline plans, it adopted an association rule mining algorithm to mine and filter alarms, optimized the outline plans according to the mining results, and produced the final plans. Experimental results showed that the method has strong versatility and fits the practical railway O&M work. It can effectively reproduce the expert O&M decision-making logic. The generated plans cover key risk points of historical alarms and significantly improve the efficiency and scientific nature of plan formulation. This method provides effective technical support for compiling O&M plans of railway information systems.