By analyzing the current railway transportation passenger ticket booking system, a prediction model of
demand for dynamic multi-seats ticket booking is built. In view of the booking demands of different classes, Poisson
distribution is employed to describe the booking arrival rates of different classes in different periods of time. Since the
practical railway ticket booking rate is hard to be available, data generated by Monte Carlo simulation are used as
numerical examples for estimation of model parameters through MLE approach. Moreover, a model for obtaining the
maximum revenue is established by dynamic programming method, and a computation with the practical data shows
that the result accords with the actual situation.