A deep learning and ensemble learning based architecture for metro passenger flow forecast

Accurate short-term forecast of metro outbound passenger flow is great significance for real-time traffic control and guidance. A good method should have high accuracy, timeliness practicality. Based on deep learning ensemble technology, this study proposes an end-to-end hybrid architecture that integrates multiple features. The innovatively bagging strategy transfer with learning, includes extensible feature processing components. In addition, presents a new coding to incorporate the operating characteristics into forecasting architecture. Use automatic fare collection (AFC) data Chengdu Metro Tianfu Square Station training verify workdays, weekends holidays. results reveal compared other widely used models, proposed in has achieved highest accuracy above-mentioned different time types. Furthermore, fusion improves model while significantly speeding up convergence training, increasing its practicability.