The Forecast of the Subway Passenger Flow Based on Smooth Relevance Vector Machine

Metro passenger flow could be taken as time series. The short-term passenger flow forecast is significant for the metro management. Many experts have researched this issue, but there also have some problems such as how to forecast the real-time passenger flow for holidays effectively. Compare with the support vector machine (SVM), relevance vector machine (RVM) takes model sparsity criterion as the priority of model weights. Smooth relevance vector machine (SRVM) is an effective method to control sparsity in Bayesian regression. It introduces an elastic noise-dependent smoothness prior into the RVM. In this paper, we use SRVM to forecast the passenger flow, and the simulation results indicate that it is effective.