With the increasing complexity of the subway system and the expansion of the operation scale, the stability and reliability of the communication signal system have become the key to ensure the safe and efficient operation of the subway. This paper uses machine learning, data mining and statistical analysis methods to build an accurate failure prediction model. The model can identify potential failure modes and warn latent risks in advance, thus effectively avoiding the failure of the communication signal system. At the same time, combined with intelligent maintenance strategies, the system can automatically locate the fault location, realize remote monitoring and quick response.