Abstract To explore the data processing of high-speed railway fault signal diagnosis based on MapReduce algorithm, the partitioning strategy of data flow was improved, and Bias classification algorithm was used to model and classify data. In MapReduce parallelization process, the data partition matrix T k was stored in line segmentation, the computing load was distributed in every node of cluster, and the time consumption of mobile data matrix and the consumption of partitioned matrix were calculated. Results show that the algorithm proposed could reduce the amount of computation in the execution process, greatly reduce the memory space consumption, and improve the counting speed in railway signal system.