Abstract: In order to solve the problems of complex environment configuration, low resource utilization and long model design cycle existing in Tensorflow framework of the deep learning platform for railway passenger marketing, this paper proposed a PAAS platform based on Kubernetes container to build a unified resource scheduling management and API access control service framework of Tensorflow container. The platform practice shows that the framework provides a reliable and stable analysis environment for passenger transport marketing, reduces the training difficulty of the model, and improves the training efficiency of the model.