Modeling dynamic pricing is very crucial for increasing revenue of airlines. People prefer to travel during holidays like summer and winter vacations, the Spring Festival, and the National Day, leading to higher demand and potentially higher flight prices. The challenge is how to leverage the temporal patterns of the calendar system for modeling the hierarchical time structures of passenger behaviors. Therefore, it is important to use a standard annual and a daily calendar system to frame temporal patterns of passenger behaviors to predict the price a passenger is willing to pay for a flight. To address this challenge, we propose a novel Session-based sequence Recommendation for Dynamic Pricing (SRDP) by utilizing calendar-airport module and passenger attributes. Our model consists of three main components: (1) passenger behavior graph embedding generation, which converts the heterogeneous features of flight and airport nodes in the passenger behavior graph into initial embeddings; (2) calendar-airport GNN module, which respectively aggregates these time and airport unit embeddings in different sessions into temporal patterns of different periodicity and spatial patterns of different airports; (3) flight recommendation and discount forecast, which predicts a passenger preference on all flights and adjusts the base discount by combining the recommendation strength of flight with the features of passengers. To the best of our knowledge, this is the first attempt to realize the dynamic pricing by using the session-based sequence recommendation system. The experimental results illustrate that our method significantly outperforms the state-of-the-art methods on one real-world dataset.