Cluster model for flight demand forecasting

Traditional models of flight demand forecasting put great emphasis on the relationship of data features between historical flights and current flights. However, these models are not suitable for the frequent changes of flight schedule, which is a situation we often face. To adapt to this situation, a forecasting model of flight demand cluster was presented. Booking procedure was defined as a vector that grew in length when near departure and multi-pass program was used in this model. Compared with regression model and pickup model that are popular in practice, the results of the model have shown its efficiency in computation speed, robustness and accurateness. This model has been applied to Xiamen Airline of China successfully and the results also have shown a good effect in practice.